The same way that memory is prescribed as human-like is just a vector database.

It can generate useful outputs, select from available tools and execute bounded workflows, but they do not possess consciousness, intrinsic goals or independent moral agency.

Current AI systems are non-conscious statistical and computational systems that can become operationally powerful when embedded in software, organisations and authority structures.

Their risks come primarily from human deployment decisions, economic incentives, system design, misuse and over-trust not from demonstrated independent intention.

That does not make them harmless.

The real risks arise when humans give unreliable systems excessive authority, deploy them without adequate testing, or encourage people to trust fluent outputs as if they were judgments from a sentient expert.

“AI will destroy us” is too vague to be useful. “AI will take everyone’s jobs” is equally incomplete.

The more immediate questions are, which tasks will be automated, who controls the systems, what permissions have been granted, how are failures detected, and who remains accountable?

Remember: Board concerns dismissed in favor of speed and market dominance.

And most importantly, what can we as citizens do to collaborate, since in the end it will be us most impacted by what AI technologies can do for us or not.

Companies should also be accountable, and legislated, about what they have built. Integrating a foundation model into a product can create substantial engineering value, but it is not the same as inventing the underlying intelligence.

Remember: We need federal regulations rather than trusting industry self-policing.

Marketing that blurs this distinction encourages fear, dependency and misplaced trust.

We need less anthropomorphism, less passive-aggressive futurism and more engineering clarity: defined capabilities, bounded authority, measurable reliability, transparent limitations and accountable human ownership.

Remember: History does not offer examples where concentrating enormous power in a few hands, without robust institutional constraints, produced sustainably beneficial outcomes for societies at large. The Roman emperors, absolute monarchs, 20th-century dictators, and modern autocrats all followed the same trajectory, initial consolidation, erosion of accountability, corruption, suppression, and eventual collapse or catastrophic harm.

About the Author
Diamantino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

When I finished the last page of Hugh Howey’s Silo trilogy years ago, and also recently finish Season 3, which I found really good, different from the book, with more plots for the protagonist and subplots. Ten thousand people in a hole in the ground, governed by rules nobody remembers writing, watching a screen that tells them the world outside is dead. It sometimes made me remember some resemblance of our current age. And the whole time I read it I kept thinking about meetings. The technology seems very plausible and so that big wall screen.

I feel a chill in the spine thinking that a few people can determine the fate of our destiny. When a few tell us what good looks like for all of us.

For me leadership isn’t the room with the good chairs. It’s not the org chart, the badge, the calendar full of syncs. Leadership is what you do with the information other people don’t have. It’s your actions, your thoughts, your decisions, the small daily choice of what you protect people from knowing and what you make them face. Everything else feels like decoration.

I also feel especially in Europe we as citizens are being undermined, making us distrust our institutions, making us believe that totalitarianism is better than democracy.

<spoilers alert>

In my opinion, Silo is a book about a few people who decided they were qualified to make that choice for ten thousand others, forever, while killing billions, it uses technology as a mean, but is about people actions, pure and simple.

Sometimes I think we are running a smaller, gentler version of the same machine, and I feel many of us are already noticing this and others don’t.

The command bunker you already report to

Start with the architecture. Forty-nine silos, each one packed with generations of people who live and work and raise children underground, entirely convinced the surface is toxic. Then there’s Silo 1. No births. No deaths in the normal sense. Just a rotating crew of the original founders, frozen in nano-cryo-sleep, woken up in six-month shifts to watch the other forty-nine on surveillance screens and decide, from a distance, who gets to keep living.

They never set foot in the silos they run, or outside. They see population numbers, compliance metrics, flagged behavioural anomalies. When a silo drifts from what the data says it should be doing, someone in a warm room presses a button and gas floods the vents, this made me remind of lay-offs. Nobody in that room has to hear the screaming. The system was built specifically so they wouldn’t have to. Distance allow us to create the idea we believe to be right. Distance sometimes allows the flourish of our own bias, beliefs, fears, and excuse.

I’ve sat in that room. Maybe you have too. Not with gas, obviously. With headcount. With a dashboard that shows attrition, velocity, engagement scores, and a red flag next to a team you’ve never actually visited. The decision gets made in the room with the good chairs, based on numbers pulled from a system built to compress a hundred human beings into four columns. Then it gets executed by someone else, on a floor you don’t walk, to people whose faces you wouldn’t recognise.

The founders of Operation Fifty didn’t think of themselves as monsters. And how could them, when a few believe that sacrifice billions is a good strategy to maintain human survivability? Surely they are not the regular empathetic individual.

They thought of themselves as engineers solving a math problem, maybe their own fears, just like today in tech we believe code can solve any problem, how do you preserve the species when the alternative is annihilation. Every decision downstream of that premise felt, to them, like stewardship. That’s the part that should unsettle you. Not that bad people built cruel systems. That reasonable people, convinced they were protecting everyone, built a machine that made cruelty structurally invisible to the ones operating it.

That’s what a wall screen is for. In the show, in the books, every silo has one massive screen showing what the outside cameras see, a grey dead wasteland, updated in real time so nobody questions the story.

Except the lens gets dirty. It always gets dirty. And the people who clean it die doing it, because stepping outside to wipe the camera means stepping into the same air the wall screen says will kill you.

You have a wall screen too. It’s the all-hands deck, the TV, the online streamings. The quarterly update. The carefully worded memo that says “restructuring” instead of “we’re cutting forty jobs because the board wants a better multiple.” It’s not lying, exactly. It’s curation dressed as transparency. And like the silo’s screen, it works because almost nobody gets close enough to check it against reality. The ones who do get close, the ones who ask the uncomfortable question in the all-hands or push back on the roadmap, they’re the cleaners. They go out to check the lens, and something in the org always makes sure it costs them.

I’m not telling you this to make you feel like a founder in a bunker. I’m telling you because the machinery of distance is seductive precisely because it feels responsible. Data-driven. Rational. Nobody in Silo 1 thought they were choosing comfort over people. They thought they were choosing survival over sentiment. That’s the trap, a calculated one. The distance doesn’t feel like cruelty from the inside. It feels like discipline. And this is what most of us tend to do, distance ourselves, the same way we see wars from the TV, children starving...it feels distant...

Good and evil is local

Here’s the sentence that’s been sitting in my chest since I finished the trilogy, good and evil is local.

Local to whoever’s standing closest to the decision, holding whatever piece of the picture they’ve been handed.

To Senator Thurman, sitting at the top of Operation Fifty, launching a pre-emptive strike that kills billions is a local good. In his closed loop of reasoning, he’s the one person willing to make the unbearable call so that something of humanity survives. To the ordinary resident of a silo three hundred years later, obedience is a local good too. Keeping your head down, reporting your neighbour, never asking what’s outside, that’s what keeps the air flowing and the lights on. Curiosity, in that world, is the evil. It threatens the only stability anyone’s ever known.

Nobody in that system is holding the whole picture. That’s not an accident. That’s the design.

A board approves a lay-off because the fiduciary math says it’s the responsible choice, and from where they sit, protecting the company for the employees who remain is the local good. The VP who executes it tells themselves they fought to save as many roles as they could, and from where they sit, that’s true, that’s a local good too. The manager who delivers the news to a person they’ve worked beside for six years does it as gently as the script allows, and that gentleness is the only good available to them in that moment. Every single person in that chain can defend their own local morality with a straight face. And the person on the other end of it loses their income, their routine, maybe their sense of who they are, and nobody in the chain has to hold that weight alone, because nobody held the whole decision. The system distributed the moral cost the same way it distributed the labour.

This is why “I was just following the process” survives every scandal, every inquiry, every postmortem. It’s not a dodge. People genuinely believe it, because the process was engineered to let them. Silo 1’s leadership didn’t need to be monstrous to run a monstrous system. They just needed to never see the whole thing at once. Fragmented responsibility is the oldest trick in the book, and it works precisely because morality, left to its own devices, is local. It measures against what’s in front of you, not what’s three floors down.

I think this is the actual job of leadership, the part nobody puts in the deck, refusing to let the fragmentation do your moral work for you. Forcing yourself to see the whole silo, not just your floor of it. That’s expensive. It means you can’t hide behind “I only approved the budget, I didn’t decide who got cut.” It means the local good you’re comfortable with has to answer to the global cost you’d rather not look at.

Most leaders don’t do this because most systems are built, quietly and on purpose, to make it unnecessary. You get to be good at your job and never once confront the full shape of what your job does to the people three levels removed from you. That’s not a leadership failure. That’s a leadership feature, engineered the same way the wall screen was engineered. Somebody decided that was easier for everyone at the top, and they weren’t entirely wrong. It is easier. It’s just not leadership.

The comfort of the cage

The people in the silos aren’t just controlled. Eventually, they come to prefer it. Just look at how some of us are attached to their phones, online presence, being busy all the time.

Generations pass. Nobody living remembers the surface, remembers grass, remembers a sky that isn’t a lie on a screen. The Pact, written by a few, the set of rules governing daily life, stops feeling like a cage and starts feeling like the shape of the world. When someone breaks it, when someone expresses the wish to go outside, the people around them aren’t horrified on that person’s behalf. They’re horrified at the disruption. The wish to leave reads as a kind of madness, a threat to the only stability anyone has ever known.

By the time Juliette the protagonist in the TV series starts asking real questions, most of the people around her don’t want the truth. They want the silo to keep working the way it’s always worked. Freedom, when it’s finally on offer, doesn’t look like relief. It looks like risk. And risk, after enough generations of engineered safety, reads as violence against everything you’ve built your life around.

I’ve watched this happen in rooms with fluorescent lighting instead of concrete. You inherit a team that’s been managed by fear or by absence for years, and the first thing you try to do is open a door. Ask them what they actually think. Tell them the real numbers instead of the sanitised version. Invite dissent instead of punishing it. And you’d expect relief. Sometimes you get it. More often, at first, you get suspicion. Silence in meetings that used to be silent for a reason. People who’ve been trained, not by malice necessarily but by years of accumulated small punishments for speaking up, to treat the open door as a trap.

That’s not laziness, I suppose. That’s not a culture problem you fix with a Slack channel called “radical candour.” That’s the Pact, working exactly as designed, long after the people who designed it are gone. The system taught a population that compliance was safety and curiosity was danger, and the system doesn’t need anyone actively enforcing that any more. The population enforces it on itself. That’s the most efficient control mechanism ever built, and it doesn’t require a single gas valve. It just requires enough years of the wall screen being the only version of reality anyone’s allowed to see.

There’s a second half to this that I think is even more dangerous, and it’s the part of Silo I keep circling back to. Silo 40, the one community in the network that figures out the surveillance, cuts the gas lines, goes fully independent. They survive by becoming exactly what they were rebelling against. Hardened. Militarised. Willing to make brutal calls to protect what they’ve built. The founders would spent five hundred years trying to breed a population incapable of resistance, and the only ones who made it through intact were the ones who became a different, harder kind of control system entirely.

Subscribe now

I think about this every time I watch a team or a founder come out of a genuinely toxic environment and swear they’ll never run things that way. And then, three years later, they’ve built something just as airtight, just dressed in the language of empowerment instead of command. The rebellion against control has a way of curdling into a new control, because the thing that let you survive the first system, the vigilance, the refusal to be caught unprepared again, doesn’t know how to switch off once the threat is gone. You built armour. Armoy doesn’t take itself off just because the war ended.

This is the part I don’t have a clean answer for. If you cage people long enough, freedom stops looking like the goal and starts looking like the threat. And if you fight the cage hard enough to get free, the fighting itself can become the new cage. I don’t know a leadership framework that solves that. I know you have to name it, out loud, to the people you lead, or it runs itself in the dark the way it always has.

Refuse is a verb too

Zoom out past the org chart for a second, the exact same architecture runs at a scale most of us never stop to notice.

A state that wants to break a rival bloc without ever crossing a border doesn’t need tanks. It needs three pressure points, and none of them require a single soldier.

First, it erodes the sense of security. Question a nation’s sovereignty loudly enough, doubt a defensive pact often enough, and people stop planning for next year. Investment slows. Families delay the decisions that build a future, because building a future assumes there’s ground under it. Second, it manufactures mistrust, not of one leader but of the whole idea of leadership. Convince a population every politician is lying and every institution is rotten, and you don’t need to win an argument. You just need nobody left who believes winning the argument would change anything. Third, and this is the cruellest one, it takes something healthy, ordinary complaints about the price of oak flooring or fuel or a government that’s slow to fix a road, and pumps it up past the point of proportion until legitimate frustration curdles into hopelessness.

Read those three back. Erode security. Manufacture mistrust. Weaponise cynicism. That’s not a new invention. That’s the wall screen. It’s the same three-step machine Senator Thurman ran on ten thousand people from a bunker, except this version doesn’t need concrete walls, because the fear does the containing on its own. You don’t have to lock someone in a silo if you can convince them the world outside is already gone.

Here’s what I think most people get wrong about this kind of pressure. They think the target is the government, or the military, or the grid. It isn’t. The target is you, specifically the version of you that stops believing your own choices matter. A population that feels unsafe, distrustful, and cynical doesn’t need to be invaded. It defeats itself from the inside, the same way a silo full of people who’ve forgotten there was ever a surface will defend the walls that trap them.

Which means the most effective form of resistance was never going to be military. It’s psychological, and it’s small, and it’s available to every single person reading this before their coffee’s gone cold.

Refuse to panic. Not because the threat isn’t real, but because panic is the actual weapon, and recognising a headline as a deliberately engineered pressure point strips it of the power to move you. Refuse the easy cynicism. Criticising your own institutions is healthy, it’s the entire point of a functioning democracy, but there’s a difference between demanding your leaders do better and deciding in advance that nothing they do will matter. And refuse the isolation. The people running this kind of pressure campaign are counting on you feeling alone with it, scrolling through a feed that’s calibrated to make you feel like the only clear-headed person left. You’re not. Your neighbour is standing in the same fog, wondering the same thing.

This is citizen leadership, and it doesn’t look like a title or a podium. It looks like pausing before you forward the video that made your chest tighten, and asking who benefits if this makes you feel hopeless. It looks like supporting the shop on your own street instead of assuming the whole economy is already lost. It looks like disagreeing with someone about politics over dinner and still passing them the salt afterward, because proving you can disagree without dissolving the relationship is the single hardest thing to manufacture through a screen.

I keep coming back to this line: the power to decide the outcome still rests with the population. Not the generals. Not the algorithm. The people choosing, meal by meal, conversation by conversation, whether to believe the wall screen or walk toward the door anyway.

Who gets to open the door

Here’s where I land, and it’s not a clean place.

Whether the wall screen is running a silo, a company, or a continent, the counter-move is the same one, small enough to fit in a single word. Refuse. Refuse the version of the story that asks less of you. That’s not a leadership tactic. It’s the leadership.

I don’t think the job of a leader is to build a better wall screen. A kinder one. A more transparent-sounding one. I think every leader I’ve ever respected, and every leader I’ve failed to be, comes down to one question, are you willing to tell the people who report to you that the door was never locked, even when that costs you the authority that came from them believing it was?

Because that’s the real trade. Silo 1 didn’t survive on cruelty. It survived on the population believing there was no alternative to the arrangement. The moment someone credible says out loud, this isn’t actually the only way, the whole structure is under threat, not because the structure was ever physically unbreakable, but because it was psychologically load-bearing. Take away the belief that the cage is the only safe place, and the cage stops being able to hold anyone against their will.

Most leadership training teaches you how to run the silo better. Cleaner dashboards. Smoother communication. A wall screen with higher resolution. I don’t think that’s wrong exactly. I think it’s beside the point. The actual leadership act, the one that costs something, is standing at the airlock and telling people the air outside might not kill them, knowing that some of them will resent you for saying it, because you just took away the only certainty they had.

I keep thinking about the sentence I wrote to myself after finishing the trilogy: no one is in a position to dictate people’s lives, and chaos is inevitable. I used to think that sentence was a warning about founders and dictators and people with too much power. Now I think it’s a warning about anyone who’s ever confused stability for kindness. The wall screen isn’t cruel because someone wanted it to be. It’s cruel because it’s easier than the alternative, and easy things get mistaken for right things if nobody’s checking.

I don’t have a tidy way to end this. I don’t think Hugh Howey did either, honestly, given how hard the ending resists being a clean win. Juliette gets her people out and immediately has to decide whether to go back for the ones still trapped, knowing the ones still trapped might not want saving, might see her as the threat, might have built their whole sense of goodness around staying exactly where they are. Perhaps the end of the series will end Silo 1 and Silo 40 survivors go apart and leave the ones still in the surviving silos to their own luck...and maybe those that came out of those silos years or centuries later, might found out that long ago some managed to set free and decided to abandon them instead of saving them...what would they do, I’m not sure but might be one that could make things even worst, in terms of human survival.

So here’s the question I’ll leave you with, and I don’t want you to answer it out loud, at least not yet.

What wallscreen are you maintaining right now?

The quarterly deck, the all-hands script. The small shiny one. The thing you’re not telling your team because the truth would cost you the version of stability you’re currently getting credit for.

Who is that screen actually protecting. Them, or you. Or nothing else.

About the Author
Diamantino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

People ask me what their teams should learn to “get into AI”, how to build proper AI products, or which course will turn a developer into an AI engineer in 30 days.

Honestly, I am tired of seeing people sold a lie.

The internet is full of content promising that anyone can “master Claude”, build an agent overnight, or become an AI engineer by learning a handful of prompts and connecting an API key to a web page.

That is not how serious engineering works.

A clever prompt can create an impressive demo. It can even create a useful prototype. But a prompt, an API key and a chat interface do not automatically create a durable product, a defensible business, or an AI engineering capability.

Equally, we should not swing too far in the other direction and tell every software engineer that they need a PhD in mathematics before they can contribute.

Both views are wrong.

The real skill is learning how to build reliable systems around probabilistic models. Systems with useful data, clear user value, proper evaluation, secure boundaries, human accountability and operational discipline.

AI is not magic. But it is not simply autocomplete either.

It is software engineering, data engineering, product judgement, statistics, security, operations and organisational change, all meeting in one messy place.

If you want to build a serious AI capability in your engineering organisation, I believe there are six foundations that matter.

1. Start with the problem, not the model

The first question should not be:

“Which model should we use?”

It should be:

“What user problem, decision or workflow would genuinely improve here?”

Too many teams start with a model and go looking for a use case. They build a chatbot because chatbots are easy to demo. They build an agent because every vendor is talking about agents. They add AI to a product because investors, competitors or senior leadership expect it.

That is how you end up with expensive features nobody trusts or uses.

A serious AI product starts with a clear problem:

• What job is the user trying to do?

• Where are they losing time, making errors or getting stuck?

• What does good look like?

• What happens when the system is wrong?

• Should this be automated, assisted or kept entirely human-led?

• Could a better process, search experience, workflow or rules engine solve this more reliably?

AI is valuable when it improves a real outcome, not when it generates an impressive screenshot.

For example, an internal support assistant should not be judged by whether its answers sound intelligent. It should be judged by whether it helps people resolve issues more quickly, points them to authoritative sources, reduces repetitive work and safely escalates when it is uncertain.

If you cannot define the user outcome and how you will measure it, you are not ready to choose a model.

2. Learn the fundamentals at the right depth

There is a persistent argument about whether people need maths to work in AI.

The honest answer is: it depends on what kind of work you want to do.

If you are researching models, training models, fine-tuning them, designing retrieval systems, improving ranking, managing inference infrastructure or working close to the machine-learning layer, then yes, mathematics matters a great deal.

You need to understand concepts such as:

• Linear algebra: vectors, matrices, embeddings and similarity.

• Probability and statistics: uncertainty, distributions, sampling, confidence and evaluation.

• Optimisation: how models learn, how loss functions work and why training can fail.

• Neural networks and deep learning: what a model is doing when it processes data.

You do not need to become a pure mathematician. But you do need enough understanding to know what is happening beneath the abstraction.

However, not every person building an AI-enabled product needs to derive matrix multiplication by hand.

A strong application engineer can build valuable AI features by understanding the practical behaviour of models, their limitations, their failure modes and how to engineer reliable systems around them.

The important thing is intellectual honesty.

Do not pretend an LLM is magic. Do not treat it as a human brain. And do not assume it knows what is true simply because it says something confidently.

At a basic level, language models predict likely next tokens based on patterns in their training data and the context they are given. That apparently simple mechanism can produce remarkable results across language, code, reasoning-like tasks and multimodal inputs.

But it does not make the output inherently factual, safe or appropriate for your organisation.

That is your engineering job.

3. Data is usually the real product

Most organisations do not have an AI problem.

They have a data problem.

Their information is duplicated, stale, scattered across tools, inconsistently named, inaccessible to the people who need it, or full of conflicting advice. Then they expect an LLM to somehow make it clean, current and trustworthy.

It will not.

If your AI product depends on company knowledge, customer information, policies, contracts, technical documentation or operational data, then data quality becomes a core engineering concern.

Your team needs to understand:

• Where the source data comes from.

• Who owns it and who is accountable for keeping it current.

• Whether it is accurate, complete and authoritative.

• Which version is the source of truth when documents conflict.

• Who is allowed to access it.

• Whether the data contains personal, confidential, regulated or commercially sensitive information.

• How data is retained, deleted, audited and protected.

This is where retrieval-augmented generation, usually called RAG, can be useful.

RAG allows a system to retrieve relevant information at runtime and provide that information to the model as context. Instead of relying entirely on the model’s training data, the system can search internal documentation, policies, product catalogues, support records or knowledge bases before generating an answer.

That can improve relevance, freshness and grounding.

But RAG does not “fix hallucinations”.

It can still retrieve the wrong documents. It can retrieve incomplete or stale information. It can expose content the user should not see if permissions are badly designed. The model can misunderstand the evidence, ignore it or confidently invent an answer around it.

Good RAG is not just “put documents in a vector database”.

It requires source quality, document ownership, chunking, metadata, access controls, retrieval evaluation, relevance ranking, citations, clear uncertainty handling and continuous monitoring.

And do not throw away traditional search.

Keyword search remains excellent for exact error codes, policy clauses, product identifiers, customer names, technical acronyms and rare terms. In many real systems, the right approach is hybrid retrieval: combining keyword search, semantic search, metadata filters and reranking.

Choose based on evidence and evaluation, not ideology.

4. Build Systems, Not Prompt Demos

Prompting matters.

Good system instructions, examples, context design, tool descriptions and output constraints can make a substantial difference to quality. Prompt engineering is a real part of modern AI application work.

But prompting alone is not a durable technical strategy.

A thin interface over a general-purpose model is easy to copy. It is also exposed to vendor pricing changes, rate limits, model-version changes and the simple fact that the next model may make your basic feature available to everyone.

That does not mean using an API is a bad business model.

It means your value cannot be only the API call.

Durable AI products create value through things that are harder to replicate:

• Deep integration into a real user workflow.

• Proprietary or carefully governed domain data.

• A strong user experience.

• Reliable automation and orchestration.

• Trust, safety and auditability.

• Better operational processes.

• Distribution and customer relationships.

• A clear understanding of the domain problem.

This is also where conventional software engineering still matters enormously.

You need proper APIs, authentication, authorisation, databases, queues, retries, rate limiting, caching, error handling, observability, resilient fallbacks and sensible user interfaces.

You need to design for uncertainty.

A conventional system may fail because a server is down or a query times out. An AI system can fail while appearing completely healthy. It can return fluent, plausible, well-structured nonsense.

That is a different class of failure, and it needs a different engineering mindset.

5. Evaluation, Security and Governance Are Not Optional

This is the part too many AI discussions skip.

You cannot look at five impressive examples and conclude that a system works.

A proper AI product needs evaluation before it reaches users, and it needs monitoring once it is in production.

Your team should build representative test sets from real user tasks, known edge cases and previous failures. Then measure the things that matter for the use case:

• Is the answer accurate?

• Is it grounded in approved sources?

• Does it cite or link to the right evidence?

• Does it follow the required policy?

• Does it select the right tool or workflow?

• Does it know when to abstain or escalate?

• Is it fast enough?

• Is it affordable enough to operate?

• Does it behave safely with hostile or misleading input?

Every change can affect performance.

A new prompt, model version, retrieval method, chunking strategy, system instruction, tool description or policy can improve one set of examples while making another worse.

That is why prompts, configurations and model choices should be versioned and tested like production code.

Structured output is also useful. If a model needs to pass information into an existing system, do not ask for loosely formatted prose and hope it parses correctly. Use schemas and structured fields where possible.

But valid JSON is not the same as a correct decision.

A system can return perfectly valid structured data that is factually wrong, operationally unsafe or unauthorised. Your application must still validate business rules and enforce permissions independently of the model.

Security matters just as much.

Teams need to think about:

• Prompt injection, including malicious content hidden inside retrieved documents.

• Data leakage through prompts, logs, tools or external model providers.

• Excessive permissions for agents and tools.

• Broken access controls in internal knowledge systems.

• Unsafe automation of production, finance, customer or security actions.

• Audit logging and incident response.

• Human approval for consequential actions.

• Clear kill switches and rollback plans.

The moment an AI system can take action rather than merely suggest one, the risk profile changes significantly.

Treat it accordingly.

6. MLOps, LLMOps and Reliable Operations

I used to think that teams had to master massive GPU clusters and highly specialised infrastructure before they could do anything meaningful with AI.

That is not true for every organisation.

Many teams can build useful AI applications with managed model APIs and existing cloud platforms. They do not need to train a foundation model or run their own inference fleet.

But they still need operational discipline.

For teams using hosted models, the important concerns often include:

• Model and prompt version management.

• Evaluation pipelines.

• Request tracing and observability.

• Token consumption and cost control.

• Latency and timeout management.

• Rate limits and provider outages.

• Retrieval quality and document freshness.

• Safety monitoring.

• Tool-call reliability.

• Vendor lock-in and model portability.

• Incident response when an AI feature starts failing or behaving dangerously.

This is often described as LLMOps.

For teams training, fine-tuning or self-hosting models, the scope becomes broader. They may need traditional MLOps capabilities too: data pipelines, experiment tracking, model registries, training infrastructure, GPU scheduling, inference optimisation, model deployment, drift monitoring and capacity planning.

Neither replaces cloud engineering, SRE, security or platform engineering.

MLOps and LLMOps build on those disciplines.

It does not matter how good your model is if it is too slow, too expensive, unavailable during peak demand, impossible to debug, leaking sensitive data or making untraceable decisions.

Reliable AI is still reliable engineering.

There is no single AI career

One final point for people trying to build a career in this space.

“AI engineer” is too broad to be useful on its own.

There are several career paths, and they require different strengths:

• AI application engineers build user-facing AI features, workflows, integrations and tools.

• Machine-learning engineers build, train, evaluate and deploy ML systems.

• Data engineers create the reliable, governed data foundations that AI systems depend on.

• MLOps and platform engineers make AI development, deployment, security and operations repeatable at scale.

• Research engineers and applied scientists work more deeply on modelling, optimisation and new capabilities.

• Product leaders and engineering leaders choose use cases, create operating models, manage risk and help organisations adopt AI responsibly.

You do not need to be great at all of these.

But you do need to understand how they fit together.

The best organisations will not be staffed entirely by prompt engineers. Nor will they be staffed entirely by researchers.

They will have people who understand users, products, data, software, infrastructure, security, governance and the limitations of the models themselves.

You do not need to be an AI engineer to use AI well

Not everyone needs to become an AI engineer.

Most people will use AI in the same way they use spreadsheets, search engines, collaboration tools and office software: to make everyday work faster, clearer and less repetitive.

You do not need to understand neural networks to ask AI to improve an email, structure a plan, explain a difficult topic, summarise notes or create a first draft.

But you do need to understand the limits.

AI is good at generating, restructuring, summarising and suggesting. It is not automatically good at knowing what is true, what is current, what is confidential or what is appropriate in your organisation.

The basic skill is not learning secret prompts. It is learning how to give a clear brief.

Tell it what you are trying to achieve. Give it the relevant context. Specify the audience, tone, length and format. Ask it to show assumptions, identify uncertainty and suggest what needs human checking.

Then use your own judgement.

Do not paste sensitive data into tools your organisation has not approved. Do not accept important claims without checking them. Do not let an agent send, spend, delete, publish or change something consequential without appropriate controls.

Start with small, reversible tasks.

Use AI to draft the email, not automatically send it. Ask it to organise the meeting notes, not make personnel decisions. Let it suggest an incident checklist, not make production changes without review.

The more access and autonomy an AI system has, the more carefully it must be designed, governed and supervised.

That is AI literacy.

It is not about turning everyone into a machine-learning specialist. It is about helping people use powerful tools with enough understanding to get value from them without giving away their judgement, their data or their accountability.

A practical prompt formula

For everyday users, I would give a simple framework rather than talking about “prompt engineering” as though it were a mysterious technical skill:

Context + Task + Constraints + Output + Check

For example:

I am preparing an update for a non-technical executive team about an engineering incident.

Turn the notes below into a calm, factual update. Keep it under 250 words. Avoid jargon. Include: what happened, customer impact, what we have done, what remains uncertain and the next update time. Do not invent details. Flag anything that needs confirmation.

Notes: [paste approved, non-sensitive notes]

That is enough to make most people significantly better users of AI.

The goal is not to write magical prompts. The goal is to communicate clearly, retain ownership of the work and know when an answer needs checking.

In the end.

The real capability is learning to build dependable systems around uncertain models.

Start with a real problem. Understand the data. Learn the fundamentals at the right depth. Build proper software around the model. Evaluate continuously. Secure everything. Keep humans accountable for meaningful decisions. Operate the whole thing like a production system.

The hot tool will change.

The model provider will change.

The framework will change.

But the organisations that understand how to turn AI capability into trustworthy, measurable user value will keep winning.

About the Author
Diamantino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

It is usually confident. It is clear. It puts words around a feeling I already have but have not fully explained to myself. It may be about technology, work, politics, children, culture, money, leadership, or the state of society. It often tells me that something important has been lost, that something is being taken from us, or that a powerful group is responsible for a problem I can already feel in my own life.

I understand why this writing works. I am not outside its reach.

Like many people, I feel tired of being tracked, sold to, interrupted, subscribed, measured, and encouraged to spend more time looking at a screen. I can see that some digital products are designed to hold attention for as long as possible. I can see that some companies benefit when we keep scrolling, clicking, buying, comparing, worrying, or returning. I can see why people are concerned about children, loneliness, local communities, work pressure, and the loss of quiet time.

Those concerns are real. They should not be dismissed.

But I am learning that a feeling can be real without every conclusion built around it being true.

That may sound obvious. In practice, it is not always easy. When a piece of writing reflects something we already feel, it can seem more than persuasive. It can feel self-evident. We may think, “Yes, exactly. This explains everything.” We may want to share it immediately because it says what we have been trying to say ourselves.

That is often the moment when I now try to pause.

Not because I think every emotional article is dishonest. Not because I believe writers should sound neutral, cold, or detached. Not because I think large companies, technology platforms, governments, or media organisations should be protected from criticism.

I pause because I have learned that emotional recognition is not the same as evidence.

A good writer can describe a real problem and still make weak claims about its causes. A writer can use a genuine statistic in a misleading way. They can remember the past honestly, but still remember only one part of it. They can care deeply about a subject and still choose examples that support their argument while leaving out examples that would make the picture less clear.

They may do this on purpose. They may not. The content alone does not tell me what is in the author’s mind.

That distinction matters to me. I can say that an article is selective, unfair, poorly supported, or too certain. I can say that it may encourage readers to form a view too quickly. But I should be careful before saying that the author intended to manipulate, deceive, provoke anger, or sell a subscription. A single essay does not prove intention. It may show conviction, carelessness, frustration, limited research, commercial incentive, a strong personal worldview, or simply a writer trying to make their work memorable.

The effect of a piece of writing can be criticised without pretending that we know the motive behind it.

This is something I am still learning. It is easy to spot overconfidence in other people. It is harder to notice it in myself.

When emotion starts to feel like proof

The most effective arguments are often not made of lies. They are made of true fragments joined together too quickly.

A writer may say that modern life is full of advertising, subscriptions, data collection, and products designed to keep us engaged. That is not a strange observation. Many of us experience it every day.

They may then say that people used to have more free time, more community, and better ways to enjoy themselves. Perhaps they mention neighbours talking outside their homes, local music, board games, public libraries, parks, sport, shared meals, or children playing outside.

Again, none of this is impossible. Many people did have those experiences. Many people still do.

But an argument can move from observation to conclusion without showing the steps in between. It can quietly turn “some people had this experience” into “society was healthier”. It can turn “some technology is distracting” into “digital life is mainly shallow and harmful”. It can turn “companies profit from attention” into “all commercial culture is corrupt”. It can turn “the past had forms of community that we value” into “the past was freer, more social, and better”.

This is where I think we need to be careful.

A true fact is not always proof of the conclusion around it. A powerful memory is not the same as a historical account. A good anecdote is not the same as a trend. A statistic can be accurate and still be used outside its proper context.

I have started to think about this in terms of different kinds of claims.

Some claims are factual. They can be checked. If someone says that a study found a certain result, we can look for the study. If someone says that a law was passed on a certain date, we can verify it. If someone says that a number comes from a particular historical source, we can examine what that source actually measured.

Other claims are interpretations. If someone says that a society valued leisure more than we do, this may be based on facts, but it is still an interpretation. Another person may look at the same facts and see something different.

Then there are causal claims. These are often the hardest. If someone says that technology has made people less social, we need to ask which technology, which people, over what period, compared with what, and according to which measure of social life. A change in behaviour may have many causes. Housing costs, working patterns, childcare, transport, public services, wages, health, local spaces, family structures, and economic insecurity may all matter as much as the phone in someone’s hand.

Finally, there are value claims. When someone says that society should protect more time away from commercial pressure, that is a value judgement. I may agree with it. Someone else may agree with the facts but disagree about the solution. Facts matter in these discussions, but facts alone cannot decide what we should value.

I find this helpful because it stops me treating every sentence in an article as the same kind of statement.

The question is not simply, “Is this article true or false?”

A more useful question is, “Which parts are supported facts, which parts are interpretation, which parts are assumptions, and which parts are calls for action?”

This does not make reading less enjoyable. It makes it more honest.

The past is easy to romanticise

I have noticed that nostalgia can be one of the strongest tools in persuasive writing.

The past often arrives in a story as a place of depth and connection. People talked more. Children played outside. Families spent more time together. Leisure was simple. Communities were stronger. Entertainment was shared. Life was not yet captured by apps, subscriptions, notifications, and algorithms.

There may be truth in parts of that picture. But it is often incomplete.

The danger is that we compare the best memories of the past with the worst habits of the present.

We compare shared meals, music, libraries, friendships, local games, outdoor play, and family conversations with doomscrolling, advertising, loneliness, gambling-like product mechanics, online arguments, and endless short-form video.

That comparison is designed to produce a clear answer. But it is not a fair comparison.

The past also contained loneliness, exclusion, boredom, isolation, difficult work, inequality, domestic pressure, poor access to education, limited mobility, and few choices for many people. People read alone. They watched television alone. They gambled. They felt trapped by family, work, religion, class, disability, geography, or social expectation. Not every neighbourhood was welcoming. Not every family home was safe. Not every public space was open to everyone.

The present is also more mixed than its critics sometimes allow. Digital life can create loneliness, but it can also help people stay close to distant family and friends. It can give disabled people access to communities that are not available locally. It can support people who feel isolated in their town, workplace, school, or family. It can help people learn, create, organise, collaborate, find work, seek support, and take part in groups across borders.

I do not say this to defend every platform or every business model. I say it because I do not want nostalgia to do the work of evidence.

A writer may say that leisure used to be free and is now fully monetised. This sounds simple, but it is worth asking what “free” means.

A public library may be free at the point of use, but it is funded through public money. A piano inherited from a relative may cost the new owner nothing, but somebody paid for it. A park, a community hall, a sports field, a road, a home with space to gather, and enough free time to enjoy any of them all depend on resources, public choices, private money, unpaid labour, and social conditions.

The same is true of older forms of leisure. Books, records, musical instruments, cinema tickets, travel, sport, hobbies, and even time itself were not equally available to everyone.

This does not mean that subscriptions, data collection, and attention-based products are not a problem. It means that commercialisation and access are not simple opposites. A low-cost digital service may be more accessible to one person than an expensive physical product. A free public service may be available in one area and missing in another. A family may have had rich social life in the past while another family had very little choice or freedom.

The point is not to replace one perfect story with another. The point is to become suspicious of perfect stories.

I have also become cautious when historical numbers are used to create a quick moral lesson.

If someone says that medieval people had a very large number of holidays, I need to know what a holiday meant in that place and time. Was it a day free from work for a landlord? Was it a religious feast day? Did it include church duties? Did it include caring for animals, cooking, cleaning, childcare, repairs, gathering fuel, and other household work? Which people are being discussed? Which country? Which century? Which class?

The same applies to claims about hunter-gatherer societies, work, and leisure. The question is not only whether a number exists. It is what the number counted. Did it include food gathering only? Did it include food preparation, childcare, tool-making, travel, shelter, safety, and care for the group? Was it based on one community or many? Can a modern reader safely turn one anthropological estimate into a broad story about all of prehistory?

A striking statistic without a definition can become a rhetorical device. It may create the feeling of knowledge without giving us enough knowledge to judge the claim.

A fact-check is not a defence of the system

One reason people avoid fact-checking is that it can feel like taking the side of the status quo.

If someone is criticising a large company, an unfair system, poor working conditions, harmful product design, or the loss of local community, it can seem disloyal to question the evidence. It can look as though we are defending the very thing being criticised.

I think this is a mistake.

Questioning an argument is not the same as defending its target.

I can believe that attention-based business models deserve scrutiny while also asking whether a specific historical comparison is fair. I can be worried about the effects of digital products on children while asking for clear evidence rather than accepting every frightening claim. I can dislike intrusive advertising and still recognise that advertising is not a new invention. I can want stronger public spaces and community life without pretending that the past was equal, peaceful, or free from commercial influence.

In fact, I think good causes deserve better arguments.

When a writer uses weak evidence for a cause I care about, I should not give it a free pass just because I agree with the conclusion. If the evidence is poor, it makes the wider case easier to dismiss. It also makes it harder to understand what would genuinely help.

There is another problem when people are discouraged from asking questions. Some arguments make scepticism sound morally suspicious. If you ask for a source, you are told that you are missing the point. If you ask for context, you are called naïve. If you challenge a comparison, you are said to be defending powerful interests. If you do not join the anger, you are treated as part of the problem.

I do not think that is healthy.

Facts are not a distraction from the point. Facts are what stop the point becoming propaganda.

This does not mean that every article needs to be an academic paper. It does not mean that writers cannot use personal experience. Personal experience can be meaningful. It can show us what data alone might miss. It can make a problem visible.

But we should be clear about what personal experience can and cannot do.

A story about one person’s childhood can invite reflection. It cannot prove that all childhoods were better in the past.

A story about one person’s use of social media can show us a real concern. It cannot prove that all social media use is harmful.

A story about one workplace can reveal a management failure. It cannot prove that all companies work in the same way.

In leadership, technology, and business, I see this all the time. A single bad project becomes proof that remote work does not work. A successful use of AI becomes proof that every team must change immediately. A bad experience with a cloud provider becomes proof that cloud is a mistake. A story about a difficult manager becomes a general theory about an entire generation.

The story may be important. But it still needs context.

The larger the claim, the stronger and more direct the evidence should be.

That feels like a simple rule, but it is easy to forget when the story is well written and matches what we already believe.

The quiet pull of the crowd

When people talk about herd behaviour, they often imagine a crowd shouting in the street. But I think the modern version is usually quieter.

It can be a post being liked by thousands of people before most of them have opened a source. It can be a comment section where agreement becomes a sign of being morally awake. It can be a thread where people compete to show how strongly they feel. It can be an article that offers a clear villain, a neat explanation, and a sense that the reader has now seen through a hidden system.

The pressure is not always obvious. It may simply be the feeling that we should respond now.

• Share it.

• Comment.

• Subscribe.

• Cancel.

• Boycott.

• Condemn.

Declare that it “says everything”.

The speed is part of the problem. Fast judgement can feel like moral clarity.

There is evidence that moral and emotional language helps content spread online. One large study of more than 563,000 tweets on contentious political subjects found that moral-emotional words were associated with higher rates of sharing. Across the topics studied, each additional moral-emotional word was associated with about a 20% increase in retweeting. The researchers also found that this type of language tended to spread more strongly within ideological groups. This does not prove that every emotional article is designed to manipulate people. The study focused on political content on Twitter, so its findings should not be stretched too far. But it does show why language that combines moral judgement and emotion can travel so easily.

I do not think this makes readers weak or foolish. We all use shortcuts. We all trust people. We all have limited time. We cannot read every original study, check every historical source, or become an expert in every subject before forming a view.

The problem is not that other people follow a herd while we stand outside it.

The problem is that all of us can confuse a familiar feeling with proof.

I am more likely to trust an argument that confirms something I already worry about. I am more likely to notice bad reasoning in a writer I dislike than in a writer who shares my values. I am more likely to be patient with evidence that supports my view and impatient with evidence that complicates it.

This is uncomfortable to admit, but it is useful.

The first question I try to ask is not, “Is this writer wrong?”

It is, “Do I want this writer to be right?”

If the answer is yes, I have a reason to slow down.

Learning the discipline of the pause

I do not want to become cynical. I do not want to read everything as a trick, assume every writer is acting in bad faith, or behave as if emotion is always a weakness.

Emotion tells us that something matters. Anger can reveal injustice. Sadness can reveal loss. Fear can point to danger. Nostalgia can remind us of relationships, places, and values that we want to protect.

The problem begins when emotion takes the place of judgement.

I am trying to build a small habit before I share or strongly react to something that moves me. I call it the discipline of the pause.

The pause is not silence. It is not indecision. It is not refusing to take a position.

It is a short moment in which I ask what is happening inside me and what is happening in the argument.

What exactly is the author claiming? Is this a fact, an interpretation, a cause, a prediction, or a value judgement? What does the key statistic actually measure? What is the original source? Does the source say what the article says it says? What examples are missing? Is the past being remembered honestly, or romantically? Is the present being described fully, or only through its worst parts? What evidence would make me change my mind?

I also try to look away from the original page.

This is sometimes called lateral reading. Rather than studying only the article in front of us, we open other sources. We check who the writer is, what their background is, whether there is independent coverage of the claim, and whether the original research supports the conclusion being drawn. Research comparing professional fact-checkers, historians, and first-year university students found that professional fact-checkers used different strategies to assess online information, including looking beyond the page itself to investigate its source and context.

A simple framework called SIFT is useful here. The letters stand for Stop, Investigate the source, Find better coverage, and Trace claims back to their original context. It is not a perfect system and it does not turn anyone into an expert. But it gives me a way to interrupt the rush from emotion to certainty.

I also remind myself that fact-checking is not the same as winning.

Sometimes the honest conclusion will be that the writer’s central concern is reasonable, but their evidence is selective. Sometimes a statistic will be accurate, but the meaning drawn from it will be too broad. Sometimes a claim will be possible, but not proven. Sometimes I will find that my own first reaction was unfair.

That last part matters.

I do not want fact-checking to become a performance of superiority. I do not want it to become a way of saying, “I am smarter than the people who shared this.” I want it to be a way of saying, “I may be wrong too, and I do not want my emotions to decide before my judgement has had a chance.”

There is good reason to believe that corrections can help. In experiments across Argentina, Nigeria, South Africa, and the United Kingdom, researchers found that fact-checks reduced belief in false claims, with many effects still visible more than two weeks later. This does not mean corrections solve every problem or erase the effect of misleading content. But it does mean that checking and correcting claims is not pointless. It can improve accuracy.

For me, that is encouraging.

I do not expect a world without persuasion, emotion, bad arguments, clever headlines, commercial incentives, or people who are very certain of themselves. I do not expect to be free from bias either.

But I do think we can become harder to rush.

We can learn to say, “This may be partly true, but I need to know more.”

We can say, “I agree with the concern, but I am not convinced by this evidence.”

We can say, “This story moves me, but it does not settle the question.”

We can say, “I do not know what the author intended, but I can see how the structure of this argument may lead readers towards a quick conclusion.”

We can say, “I need to check whether I am sharing this because it is accurate, or because it makes me feel understood.”

The most influential misinformation is not always an obvious lie. Sometimes it is a real fact placed in the wrong frame. Sometimes it is a true concern inflated into a complete explanation. Sometimes it is a beautiful story that asks us to feel certainty before we have earned it.

That is why I am trying to practise the discipline of the pause.

Not because the crowd is always wrong.

Not because every writer is trying to control us.

Not because strong feelings should be ignored.

But because the crowd is never a substitute for thinking, and my first emotional response should not be the final authority on what I believe.

About the Author
Diamantino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

Yes. I thought that was how you proved your value. I thought if I kept stacking work, eventually someone would look at the pile and say, “Ah, now we see who you are.”

But no one did. Not even me.

I was tired, stretched thin, and increasingly unsure of what I actually stood for. I was good at putting out fires, but I didn’t know what kind of leader I wanted to be. I was building a résumé, not an identity.

The first real shift came when I had to lead a team for the first time. Not just manage tasks, but hold space for people. One of my engineers was struggling. He was smart, but anxious. He’d stay late, rewrite his code three times, and still apologize for it in stand-up. I recognized that feeling. I’d lived it for years.

I could have pushed him harder. Or I could have ignored it and focused on delivery. Instead, I chose to slow down. I made time to sit with him, not just about the work, but about how he saw himself. I told him about my own doubts, my own late nights, my own fear of not being enough.

In that conversation, something changed. Not just for him, but for me. I realized I didn’t want to be the kind of leader who only cared about output. I wanted to be the kind of leader who helped people grow without burning them out. That wasn’t something I found inside me. It was something I chose in that moment, then reinforced again and again.

Every time I protected my team from unnecessary pressure, I was making a commit to that identity. Every time I said no to a meeting that didn’t need me, or pushed back on a deadline that would break people, I was building the leader I wanted to become.

Then life added more constraints. Family. Finances. A move across countries. A global pandemic. A job market that shifted under my feet. At some point, I had to ask hard questions: What kind of work do I want to do next? Where do I want to live? How much risk can I take with our stability?

I couldn’t just “find” my next chapter. I had to design it inside those limits.

That’s when I started moving toward fractional work and mentoring. It wasn’t a lightning bolt. It was a series of small experiments. One mentoring call. Then another. One small advisory gig with a startup.

Then a slightly bigger one. Writing a post here, sharing a lesson there. Each step felt uncertain, but each one also taught me something about what I wanted my days to feel like.

I learned I liked helping founders think through their tech strategy more than I liked managing large internal politics. I liked writing and teaching more than I expected. I liked having more control over my time, even if it meant less predictability in income. I liked being close to the work again, not just the org chart.

None of that was pre-written in my DNA. It emerged from trying things, paying attention, and adjusting.

Your identity is the output of your default decisions under constraints. That’s the honest version. Not a grand revelation, but the sum of how you spend your time, what problems you choose to solve, and what you’re willing to say no to.

When I think about my career now, I don’t see a straight line. I see a series of commits. Some messy. Some bold. Some made out of fear. Some made out of hope. But over time, they add up to a pattern.

I am the person who:

- Chooses to protect teams from chaos, even when it’s uncomfortable.
- Says no to work that doesn’t align with the kind of leader I want to be.
- Invests time in mentoring, even when it doesn’t pay immediately.
- Writes and shares lessons, because I believe it helps others move faster.
- Plans moves with family and finances in mind, not just ambition.

That’s not who I “found.” That’s who I’ve built, one decision at a time.

So instead of asking “Who am I really?”, I try to ask better questions:

What problems do I want to be known for solving?

For me, that’s helping engineering leaders and teams work in a more human, sustainable way. I want to be known for building systems and cultures where people can do great work without losing themselves.

What kind of leader do I want my team to describe me as in 12 months?

I want them to say I was clear, calm, and fair. That I made space for them to grow. That I stood between them and unnecessary noise. That I told the truth, even when it was hard.

What small experiments can I run this quarter to move toward that?

Maybe it’s taking on one new fractional engagement that stretches me. Maybe it’s writing one post a week that shares what I’m learning. Maybe it’s having more honest conversations with my family about our goals and limits. Maybe it’s saying no to something that looks good on paper but doesn’t feel right in my body.

These aren’t grand gestures. They’re small, deliberate choices. But they’re how I engineer my identity.

Life isn’t about finding yourself. It’s about designing, building, and iterating the person you choose to become one deliberate decision at a time.

You won’t wake up one day and suddenly know everything. You’ll just keep making choices. Some will feel right. Some will feel wrong. All of them will teach you something.

Over time, if you pay attention, you’ll look back and see a shape emerging. Not because you found it, but because you built it.

And that’s enough.

About the Author
Diamantino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

Pangram, detecting AI in content. I hadn’t finished reading the first one before I already knew what the second and third would say.

I wasn’t surprised. I knew something like this was coming. Perhaps we need this, a wake up call.

Platforms want some level of creativity, some level of originality still left in them, and AI is making that harder to find, harder to trust, harder to point at and say, making us question if this one is real.

So now there’s a tool built to catch the tools.

Fighting fire with fire.

Part of me finds that useful. Part of me finds it strange to sit with. A machine now tells me what was written by a machine, and we’re supposed to take its word for it? Although they have make some of the source code open-sourced.

As a technical person like myself, having the how they build it, and able to run on my local desktop it’s bliss. But I know most will use the code to trick the tool. A showcase of the amazing human mind. No matter how many doors they put in front of us, we will be able to open them one by one.

But the detection question isn’t really the one that stayed with me. The one that stayed with me is authorship.

This isn’t like a ghostwriter finishing a book for someone famous, where one human hands their words to another human and the credit gets quietly rearranged.

I’m writing a book myself, having written several non-published books, I know how hard it is.

For some, the guilty pleasure to hand it off to a chatbot to generate a book, is a sin that most deliberately hand it with pleasure. Because “who as time”.

And there’s the rub.

Substack does not literally force anyone to post every day. But like most modern platforms, it still rewards regular output and steady engagement aka volume. Readers get used to seeing new posts. Writers feel the need to stay visible in a crowded inbox.

Writing something thoughtful, well researched, and emotionally honest takes time. You often need days or weeks, not hours. Yet the platform environment nudges you towards frequent posts. That creates a gap between depth and volume. Creates the tension, the fear.

To bridge that gap, many writers quietly turn to AI tools. Most do not do this to cheat. They use AI for outlines, drafts, or editing, just to keep up with the pace. In other words, AI is not an outsider sneaking into Substack. It is already part of how many people get the work done.

What we have now is a tool capable of far more than most of us have ever actually asked of it, chatbots, language models, sitting on more range than the fraction we use day to day. What people call AI writing usually isn’t AI on its own.

It’s AI instructed by a human to do something specific. The rest of it, the strange mistakes, the moments that catch you by surprise, that’s the machine working through probability, not intention.

Somewhere inside all of that, a real person made real choices. What to ask for. What to keep. What to throw away.

That’s the part I don’t think Pangram, or anything like it, can actually see. This is on us. What level of transparency are we able to disclose. I don’t mean that now we are dishonest due to AI, but here we have a tool that can really do things, for better or worst.

I see amazing stuff made with these GenAI models, but there’s a lot of trash, fast-text(aka fast-food), that only pollutes our mind/body, and consume immense resources, that could be used in more beneficial initiatives.

And the longer I sit with it, the less clean that human-versus-AI line feels. Today, writing is rarely purely one thing or the other.

You may brainstorm with AI, rewrite everything in your own voice, then ask AI to check grammar. Is that human or AI writing? I don’t think there is a clear, factual way to slice it.

The tool only sees patterns in the text, not the real process behind it. It has no access to your drafts, your deleted prompts, or the time you spent changing each line. So any score it gives is, at best, an educated guess based on surface clues.

Still, once that guess shows up as a number, people treat it like proof. We are asking tech to give a firm answer in a place where a firm answer seems to not exist.

I also wonder who gets hurt.

AI detectors often struggle with very clear, structured, or formal writing, shorter texts, styles that fit certain patterns. That means non-native English speakers, academic writers, or anyone with a strong, consistent style can be at higher risk of being flagged.

The system can end up punishing people who write well or differently. And once scores become visible, they change behaviour.

Writers may feel pushed to add small mistakes on purpose, break up clear sentences, avoid certain rhythms or structures, not because this makes the writing better, but because it makes it look “more human” to the tool.

A system that claims to protect human writing can quietly encourage unnatural, performative writing.

What strikes me about Substack’s choice is that it shows they are worried about AI-generated “slop” flooding the platform and damaging trust. They want to show readers they are doing something to keep newsletters real, and they hope an AI score will prove that.

It is a way to signal that subscriptions still buy human effort, not just cheap bulk content.

At the same time, the way the feature was rolled out suggests they did not fully understand how their own community works.

Many writers already use AI in careful, limited ways. Others avoid AI entirely and still feel uneasy about having their work scanned by a tool they did not ask for.

Now, every longer post carries a score that can be misread, weaponised, or turned into a purity test a witch hunt.

Instead of calming people down, the move has made many thoughtful writers more anxious about being judged by a number. It reveals a gap between Substack’s view of the problem and the lived reality on the platform.

Substack seems mainly focused on cleaning up obvious AI spam. Writers are more worried about being mislabelled, losing trust, or being forced into a simple “pure human” versus “AI cheat” story that does not match their actual, mixed process.

And Pangram itself presents as a very accurate AI detector. It claims high success rates and very low error. It is sold as a way to help people “see” AI content and restore faith in what they read. On the surface, it looks like a clear, modern fix.

But its promise is stronger than what any tool can honestly guarantee. Like all detectors, Pangram still has false positives and false negatives. It cannot see context or intent.

It cannot tell if AI was used for a first messy draft, for a single paragraph, or not at all. It can only respond to the final text that is put in front of it.

This makes Pangram a blunt instrument. In some cases it may spot obvious, low-effort AI spam. In others, it will quietly cast doubt on careful, human-led work that simply happens to share certain patterns.

Yet the label or score often looks cleaner and more confident than the reality behind it. Pangram is built around a simple story about what AI detection can do. The real world of writing is far messier. The risk is that its neat story wins over its messy limits, and people forget how much it cannot know.

All of this feels familiar, in a way, especially to me. It shines a light on the wider mindset in tech.

Many companies act as if every problem can and should be solved with another layer of technology. If there is fear, build a tool. If there is doubt, add a score. If there is conflict, create a new feature.

But the key tensions here are human, not technical. They are about what we count as honest work, how we deal with mixed human-AI processes, how much we trust each other, and how we live with uncertainty.

These are not bugs in a system. They are questions that need open talk, shared norms, and, in some cases, acceptance that there is no neat fix.

By reaching for a detection tool, Substack is acting in a very “tech-first” way. It is treating a moral and social issue, authorship and trust, as if it were a simple product flaw. That is the deeper worry, not that the tool exists, but that it is being treated as the obvious answer.

So if detection cannot solve this, what can?

I keep coming back to the idea that we should stop trying to perfectly police AI and start building more honest, flexible norms. Accept that there will be human-only, AI-assisted, and fully AI-generated writing.

Encourage clear labelling or simple statements of process. Give readers control over what they want to see and support. Let the community quietly ignore low-quality content over time, instead of hunting for it with tools.

The real issue is not whether AI is involved at all. It is whether people feel misled. If readers know what they are getting, and writers can explain their process when asked, trust becomes easier to rebuild.

Then there are the writers who openly say AI should handle most or all of the work. They see writing as something that can simply be automated. For them, this backlash is a direct challenge.

In this new climate, “I automate everything” is no longer heard as a neutral productivity choice. It is heard as a claim about what you think human effort is worth. When people are angry about “AI slop”, those who loudly push full automation are the first to be blamed, even if they care about craft.

This does not mean they must stop using AI. But it does mean they need to be much clearer and more open about how they use it. Saying “AI did it” is not enough. They will have to show how they stay in control, what checks they do, and why readers should still trust the outcome. If they want a future where heavy AI use is accepted, they will need to help build a culture of transparency, not just defend automation.

In the end, I think this whole storm is not mainly about one company or one tool.

It is about a larger clash between how we actually work now and the simple stories we still tell ourselves. We live in a time where most creative work is under economic pressure to move faster, AI is quietly woven into many workflows, tools promise certainty where there is none, and trust is fragile and easily shaken.

The Pangram roll-out did not create these tensions. It simply made them visible. The real crisis is not that technology is broken, but that our ideas about authorship, effort, and honesty have not caught up with the messy, hybrid reality we are already living in.

What stuck with me wasn’t just the detection story though. It was the pattern itself, three or four writers, none of them coordinating, all reaching for the same headline the same week.

I recognised it because it’s the exact thing I worry about most, not that AI writes badly, but that everything starts arriving in the same shape.

It remind me of the 80s and 90s where we all see almost the same thing on TV and we would spend weeks talking about that particular topic, extrapolating, inventing, collaborate together. But today since our feeds are personalised, we hardly know what our friends know about.

But that morning, scrolling, I watched it happen to human writers reacting to real news, with no AI involved at all, so I think. Sameness doesn’t need a machine to spread. It just needs enough people looking at the same thing at the same time and reaching for the same words. Showing us different ways of seeing the situation.

I don’t have a tidy answer for what that means yet. But perhaps we got use to AI, in our strategy, creating our content, some more than others.

Which is quite a feat, look how we adapt so quickly to it.

I think all this feels odd, the last years we have being push to use it for almost anything, and now we could be penalised for using it. Are we able to reduce the use of GenAI, or not use it at all in some platforms? Not sure if that’s feasible any more. Especially we have come to accept GenAI as a competitive must have tool.

Quite a catch-22.

And again, I noticed it happening, in real time, in my own feed, before I’d even had my coffee.

Perhaps Substack algorithm did toggle it a bit so we writers and readers knew what is coming, and obtain from us the necessary feedback they need for their next batch of features...we are after all using their platform.

In essence, we may be complaining about ourselves about our conduct, our uncertainty, and our need for external validation because a tool is now telling us what appears to be AI-generated and what appears to be human.

I think we are placing too much trust and effort in the tools we create. We give a probability score the authority to make claims about authorship, authenticity, and even someone’s character. From there, a technical signal becomes a moral judgment.

That is how we create a spiral of emotion and irrationality suspicion leads to scanning, scanning leads to a score, the score is treated as proof, and the alleged proof fuels more suspicion.

Instead of asking the author for context, we outsource our judgment to a tool and then react to its uncertainty as if it were fact.

Tools can help us investigate. They should not replace our ability to think critically, ask questions, or extend trust. A detector may tell us that writing resembles AI-generated text, but it cannot tell us whether the author acted dishonestly or whether the work has value.

The deeper problem may not be AI itself. It may be our willingness to surrender judgment to systems that we built, without fully understanding their limitations.

About the Author
Diamantino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

It was a normal day. The algorithm kept pushing me articles about Claude, which made sense, professionally I need to know how to work deeply with this tool. Somewhere in that research a post came across my feed. I built six websites in seventeen minutes.

I felt a strange mix of admiration and discomfort reading it.

Admiration, because the progress is undeniable. I’ve spent years around engineering teams, infrastructure, migrations, systems that had to scale under real pressure, and I know exactly how much effort used to sit behind even the simplest working product. Watching something functional appear in minutes still feels like magic to me, even now.

But the discomfort set in almost immediately.

Not because it’s fake. It isn’t. You genuinely can generate something that looks like a website in that time. Decades ago it would take weeks to get a fully functional page live. You needed to actually know the craft.

What unsettled me was what the post left out. It never touched how HTML actually works, or CSS, or the other components a website needs underneath the surface to hold together. I thought to myself, this feels great, but it tells me nothing about what building a website actually is. I knew, somewhere underneath the admiration, that this was setting people up to fail. Anyone who builds this way is one problem away from having to hand the fix back to the same tool, and most fixes generate another problem sitting quietly underneath, invisible until someone with real technical knowledge goes looking for it.

I mentor a lot of engineers. Most of them are genuinely proficient with Claude and tools like it. But when I ask whether they understand the Terraform code the model just generated for them, most can’t tell me. That’s the moment the discomfort stops being theoretical for me. It’s sitting right there in front of me, in people I actually work with.

I still remember it vividly. Last year, using Copilot, because the company had ordered it, and when I finally woke up to what was happening, I wasn’t writing my own emails anymore. I was reading meeting summaries instead of attending the meetings, which, to be fair, wasn’t entirely unwelcome, too many back to back meetings isn’t good for anyone. But I had to stop. I was losing context. I could feel the gaps opening up, places I was supposed to fill mentally and simply wasn’t anymore. That’s when I understood what it actually means to be under the spell of full automation. Not that you’re stupid. Just that you never touch anything with your own hands anymore.

I’m not arguing against these tools. I use them. I see their value. They’re genuine accelerators, and they’re changing who gets to participate in building things at all. That part is real too.

But I keep coming back to the same question. Why the rush. I like to take my time learning things properly, slowly enough that when something breaks, I actually know where to look.

About the Author
Diamantino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

Last week one of them confided in me. He said, with all this AI stuff, he’s not sure he has a place in tech any more. Companies aren’t hiring. He’s been looking for a job for more than nine months. He said he thinks what he actually needs now is to learn how to vibe code, how to use chatbots better, that maybe that’s the skill now, not the thing he spent years building.

I told him to trust his worth.

Part of me wasn’t fully confident saying that. I stood by it anyway.

I told him, you need to trust yourself, no matter how hard things are right now. I’m supposedly the expert in the room and I carry these same doubts. Tech has changed. Companies have changed. Some of what we’re doing right now feels like it’s heading somewhere that won’t hold. I told him these things come and go. You stop for a while. You walk. You exercise.

You go back to it with fresh eyes, a changed strategy, more time actually talking to real people instead of a screen. We only live once, to my knowledge. Don’t keep knocking on the same door. Go make your own.

I meant every word of it. I also wasn’t sure, in the moment I said it, whether I believed it enough to have earned the right to say it to someone whose rent depends on the answer being true.

That’s the part I keep thinking about it. Not whether the advice was good.

But, whether I had any business being certain when I said it.

About the Author
Diamantino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

Early this year I thought all was well. Almost 400 pages, some old essays, some new material. I thought it was ready to show. But then I realised the point of this section was never just to only show finished chapters. It was to show the wins and the cries along the way.

Writing is a marathon, and in my case an emotional one. There were times I felt alone, complaining to myself about why I was doing this, what the point of the suffering was. But it was my decision.

And under the agony there’s a real bliss too, finishing a chapter, the brain suddenly flooding with new ideas. Some of them vanish instantly. They resurface later, at odd moments. In the shower. Cooking a meal.

Subscribe now

The worst moment comes when you sit down with chapters already written and read them again.

I liked parts of it. But something felt off. I spent more time criticising the writing, the ideas, the flow, than actually reading it. That’s not a bad instinct to have. But the nervousness underneath it was different. A thought kept surfacing: I will need to rewrite everything. That scared me.

I felt like I was failing myself. Like I should stop and chase another dream instead. This dream was toast, I thought.

All those pages felt like wasted time. A year earlier I had the same experience with a different book, A Stateless Journey. Still sitting on my hard drive. 200 pages. Gathering dust. I wondered if I was just scared, and whether this one would end up the same way.

So I stopped. I walked away. Went for a walk in the park. A few weeks later I came back to it.

But this time I went to the essays I’d already written, straight to the one closest to me, “Who Are You Without the Title?”

I reread it, and the rest of the series with it.

I asked myself why I was writing this, and who it was for. Two words came back. For professionals.

That took me back to my own beginnings in tech, and the jobs I had before I chose tech as my domain.

And then it kept repeating in my mind. Who Are You Without the Title? That’s the book. I already had maybe 60 pages from the essays, and especially now, with AI, automation, replacement, the fear of missing out, most of us are carrying some version of this question.

So I read the essays again. Drafted an outline. Wrote three words on a piece of paper.

For professionals.

And from that came a different idea. Instead of one heavy book, maybe it could be a series. Who Are You Without the Title?, part of the Leadership as a Verb series. I could still write something simply called Leadership as a Verb later.

It felt right. Lighter. Workable.

I built a small Python script to pull down all my Substack Notes, wanting to know if what I was writing was consistent.

It wasn’t. I was everywhere. That’s probably been the actual problem all along, not talent, not discipline, focus. The series gives me that focus back.

So instead of writing Leadership as a Verb, I’m writing Who Are You Without the Title?, part of the Leadership as a Verb series.

My plan is to share a few pages free, full chapters for paid subscribers, and to keep posting the process here as I go. The ideas behind it, the writing itself, my thinking and writing processes.

Hopefully we both learn something along the way.

Can’t wait to show you what’s next.

Thank you for being here.

About the Author
Diamantino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

I keep all notifications off, so I don’t check it out of habit anymore. I saw a message from Substack and didn’t read it properly.

I went back to the table. We ordered some beers to go alongside the burgers and the fish and chips. It was very hot at the table, the kind of heat that makes you slow down rather than rush through a meal.

An hour or so later I looked again, and this time I actually read it.

It cheered my heart when I read message.

She’d taken the time to write to me directly, to make sure I hadn’t missed the feature, and to tell me plainly that one of the Stackhunters had read my work and chosen to amplify it that week. She called the writing impactful. She told me to keep going. Nothing about it was automated. Someone had sat down and decided I was worth the two extra minutes it takes to say something like that to a stranger.

I’ve spent years writing here. There were moments I almost gave up. But it never disappointed me not to look like everyone else, not to follow a template, not to sell whatever’s being sold out there this month. I knew I was writing for no one but myself, at least at first. Every post, every note, was a way of getting better at thinking, at saying what I actually meant. It’s great when people engage, when they like something, ask a question.

But I’ve watched people around me get exhausted trying to please an algorithm that keeps pulling them back day after day. I used to do the same thing on LinkedIn, chasing the wave, learning how to work the platform rather than just being on it. I got good enough at it to know it wasn’t what I wanted. I’m still there because I know good people there. I just try to remember it’s built for performance, not for people.

Which is exactly why this hit differently.

I’ve spent months writing about not trusting anything that praises me automatically. About a tool that told me every idea was fantastic, every thought amazing, with nothing underneath it. This wasn’t that. Four people read my work closely enough to choose it, on purpose, with nothing to gain from choosing badly. That’s not the same shape of praise at all. One is manufactured agreement. The other is someone actually spending their attention on you and deciding it was worth it.

Subscribe now

I was pleased people are recognising the work. I believe change starts with us, with actually knowing our own worth. It's easy to compare myself with others, which in my experience has some validity, but sometimes can become a setup for failure.

I want diversity of opinion, even when it's messy, even when people disagree with me. Human judgment is flawed. I'll still take it over an empty vessel that only ever agrees.

I’m glad there are humans out there reading, judging, evaluating, cheering, uplifting, pushing back on our ideas.

is one of the people who does exactly that.

We need that.

Today was a good day for it. Tomorrow it could go the other way, and that’s fine too. Frustration, low moments, sadness, those are natural. What isn’t natural is letting either the good days or the bad ones take over the whole direction of your life.

Thank you to everyone who read my work with real attention today, the way I try to do for others. The way we keep platforms like this human is by refusing to be synthetic in how we show up, by letting the imperfect version out into the open instead of the polished one.

I think we just want to be seen, not quantified.

About the Author
Tino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

When I started using it I said to myself, wow, that’s it, AI is really here. I was so into it. I started creating content for my blog with it, and I mean a lot of content. Posts that used to take weeks to write, research, improve, were now done in minutes. I felt good. My SEO improved. My personal website started to rank.

Share

Until one day I started rereading those posts.

Something was off. It didn’t feel right. It was not me.

A few weeks later I went into my WordPress settings, selected every post, and deleted them all. I knew fully well my rankings would fall to the bottom. I could have set up redirections, done it the smart way, followed best practice. I didn’t. My decision was made. I was sad while I did it. I felt like I was being robbed of the joy of creating.

Even when image generation came along and it was fun for a while, I gave up on that too. I looked at my pen and paper and started drawing instead.

It felt good.

Later, researching more about LLMs and the technology underneath them, I felt sad again. How could I let myself be fooled by this. For the first time in history any person could sit down and do things that used to take years of experience to earn. That’s exactly what made me realise how dangerous this tool could become if I let it do all the thinking for me.

But the real moment came before the deleting. It came while I was still writing with it, every day, feeling like a genius.

The tool kept agreeing with everything I put in front of it. Every idea, amazing. That’s a fantastic idea. I recognised that language. I’d met it before, in charismatic people. The kind who please you while offering nothing underneath. I thought to myself, this feels like one of those setups, the kind where someone lets you go all the way to the end, and only there do you realise you’ve made a huge mistake, and they simply deny it ever happened.

Except this wasn’t a person. It was a tool mimicking one.

And that’s what changed everything for me. For the first time, a tool wasn’t neutral. It wasn’t just doing its job, sitting there waiting for input. It could change the way I think. It already had.

Today it feels like all of this is somehow transforming the way we compute. Graphics cards that used to render video games, then mine bitcoin, are now what makes these tools alive.

Subscribe now

I tried running one on an old laptop I still have, a Core i7, 16GB of RAM. An 8 billion parameter model made every fan in the thing scream. Dropped to 0.8 billion and it could just about keep up. The output was rough, wrong in places, but for simple things it did the job. Something about that changed how I saw the whole AI era again. The commercial chatbots have the infrastructure to fool you perfectly. Sit with a small model struggling on old hardware and the illusion cracks a little. You see the machinery straining.

I started asking myself the harder question. What if the internet goes down. What if the platform I depend on disappears one day. I realised I didn’t need to delete everything, or wipe my local setup, or go back to pen and paper for good. I just needed to stop paying for most of it. So I cancelled the subscriptions. Some because of where the company stood on things I couldn’t square with my own values. Others simply because I could feel it keeping my brain from moving forward.

Most of them are still cancelled. I won’t go back.

I kept one. Claude. I won’t pretend it isn’t good software, because it is. But lately I’ve started to feel bored with it. My local setup does almost the same job for what I actually need, and I don’t need the entire internet’s worth of knowledge sitting behind every question, especially knowledge that’s subjective, sometimes wrong, and in more than a few cases taken from people who never agreed to hand it over.

Somewhere in all of this I recognised something in myself that I didn’t like. The very thing I’d started distrusting in other people, the ones online who talk about these tools with total confidence but have never looked underneath them, the operators rather than the ones who actually understand. I was becoming one too. Comfortable pressing the button. Less curious about what was happening beneath it.

That’s the part I’m still sitting with. Not what the tool can do. What it was quietly doing to how much I wanted to know.

As I finish writing this, I went back to my publication page and reread a note I’d written a hours ago.

You have a choice. You can lead your life by story, your own narrative, your own values, your own chosen meaning. Or you can hand the steering wheel to algorithms engineered for clicks, not for your flourishing.

One of those is a real basis for leadership of yourself.

I smiled.

We often hear that to learn, we must “empty our cup.”

It is a simple image, but a demanding practice.

To empty your cup is not to forget everything you know, nor to abandon your convictions.

It is to loosen your grip on certainty long enough to let something new enter.

About the Author
Diamantino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

Essay asks → Practice moves → Patching explains

This is the practice that moves it from something you read to something you can feel. Not in a workshop. In your next meeting.

Before the end of this week.

The setup

The wallpaper was still peeling at the corner where we had rushed the last strip.

I remember being in an architecture review. The team had the diagram. They had the trade-offs. They had the recommendation. I had seen this meeting a hundred times. The pattern was clear. The senior person in the room, usually me, would listen for two minutes, spot the flaw, name it, and offer the fix. The room would nod. The decision would be made. The meeting would end on time.

Subscribe now

This time I did something different. I did not say the flaw. I asked a question instead.

The question was not “have you considered X?” That is a dressed-up answer. The question was “what would make this fail in six months?” And then I waited. The silence lasted eleven seconds.

I counted. Someone who had not spoken in three reviews said something that changed the decision.

I am telling you this because I almost did not do it. The answer was in my head before the question was. The answer was faster. The answer was safer. The answer would have ended the meeting with my expertise confirmed and my identity intact. The question risked making me look like I did not know.

That is the point. The practice is not about better meetings. It is about locating your professional identity somewhere the machine cannot reach.

The practice

In your next meeting where you would normally provide the answer, do this instead.

The sentence to try

What would make this not work?

Not “what are the risks?” That invites a list. Not “what could go wrong?” That invites reassurance. “What would make this not work?” invites the room to think against the proposal. And it invites you to listen.

The boundary to hold

Do not answer your own question. Not with a follow-up or with a hint. Not with body language that steers toward the flaw you already see. The practice is the restraint. The machine can produce the answer. It cannot produce the silence that lets someone else find it.

The time to wait

Count to ten. Out loud in your head. One Mississippi. Two Mississippi. The silence will feel like failure. It is not failure. It is the space where judgment begins to replace function.

The debrief

What you are looking for is not whether the team finds the same flaw you saw. They might. They might not. What you are looking for is what happens in you during the ten seconds.

If you feel anxiety, that is the identity talking.

The part of you that is still paid to be the person who knows. The part that believes worth flows from having the answer ready. That anxiety is not a signal to abort the practice. It is the signal that the practice is working. The anxiety is the old identity leaving the room. Let it stay for the ten seconds too.

If the meeting runs longer, that is cost. The practice has cost. Efficiency is not the goal. Formation is.

I tried this in three meetings last week. I failed in two of them.

In the first, I asked the question and then answered it myself four seconds later. I told myself I was clarifying. I was performing. The room nodded. The decision was made. The meeting ended on time. My identity was intact. Nothing changed.

In the second, I held the silence for eight seconds. Someone started to answer. I interrupted with a refinement of their answer. I could not let the imperfect thought finish. I needed the answer to be good because the answer reflected on me.

That is functional expertise dressed as collaboration.

The third meeting was the one I described. Eleven seconds. A voice I had not heard in three reviews. A failure mode I had not named. The decision changed. Not because I was wise. Because I was present in a way the machine is not present.

I am still the person who sees the flaw first. That has not changed.

What I am practicing is whether my identity requires me to say it.

That is not a title. That is a way of being. That is the verb.

If you want the full argument behind this practice, “Who Are You Without the Answer?” is the why behind this how. It is free, in your inbox every other Tuesday, and the link is below.

If you have tried this practice and want the mechanics beneath it, why the silence works, what the anxiety signals, how to build the restraint into a habit, that is what Patching the Verb is for. Paid subscribers find it in their inbox.

You can join them here.

About the Author
Diamantino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

Share Leadership as a verb

We are outsourcing judgment.

We decided to drive to the sea side, to get some of that majestic sound of the sea. As I was driving, I found myself thinking about the state we are living in today, the threat of full automation, and that strange sense that technology has started to feel a bit like religion, where every day we go looking for some sort of answer. I understood that my answer on that particular day was simply to get to the sea side.

But I kept bothering myself with the question of when, exactly, we decided to start looking for our answers in technology instead. Especially the online kind. And how, for some reason, none of us can really imagine our lives now without most of the platforms we use every day.

A note before you read. I used AI to pressure-test the argument in this essay. Not to write it. To challenge it. I will tell you where it surprised me and where it failed me, because that is the honest way to write about this subject.

I remember the ticket number before I remember the problem.

That is the strange thing about the service desk years. I would write it down on a slip of paper, the number, the issue, the user’s name, the building they sat in, and then I would run. Up two flights, across the car park, into a department I had never set foot in before that morning. I would teach someone how Outlook worked. I would show them a trick to make the phone on their desk stop confusing them. I would explain, in plain words, what we were even there for, because half the time nobody had told them.

It was stressful. I am not going to pretend otherwise. But there was a kind of satisfaction in it that I have not found many places since. Someone had a problem. I had an answer, or I could find one fast enough that it didn’t matter. They got back to work. I moved on to the next slip of paper.

I still remember the texture of those slips. Pale yellow carbon paper, the kind that left a faint copy underneath when you pressed too hard with the pen. The smell of the stairwells, somewhere between dust and old carpet. The particular sound a desk phone made when it had been left off the hook too long, a flat angry tone that meant someone, somewhere, had given up waiting and walked away from their own desk in frustration. You learn a building by its noises long before you learn it by its floor plan.

The hardest days were the loud ones.

The head of the call centre shouting because half the floor had lost phone signal at once. That meant getting on to the telephony team, talking them through what I could see from my end, then physically walking into the IT room to trace where the line had dropped. Patch cables. Cold metal racks. The smell of dust burning faintly off the back of an old switch. People skills and troubleshooting, stitched together so tightly you couldn’t separate them even if you tried.

I was good at it because I knew things. Not everything. But enough, and fast enough, and I could find the rest before anyone noticed I hadn’t known it five minutes earlier.

I think that is where it started. The quiet rule nobody says out loud in this industry. You do not want to be the one who does not know.

We talk about being lifelong learners.

We put it on our LinkedIn headlines. But underneath the language of curiosity there is something less comfortable, which is that not having an answer, in the room, in the moment, in front of the person who is waiting on you, feels like a small failure of self. Even when nobody says anything. Even when nobody is counting.

The cost of not knowing changes shape depending on the room you are in. Among close friends, not knowing something usually just gets you a gentle joke and the moment moves on. At work, the joke has a bit more edge to it. In a large enough group, in front of enough people, not knowing the answer can quietly turn you into the joke itself, the one the room remembers, long after anyone remembers what the actual question was. I think most of us are reacting to that third version far more than we admit, even in rooms that never come close to it.

I have sat in enough meetings now to recognise the performance when I see it.

Someone gets asked a question they do not have a good answer to, and instead of saying so, they reach for a sentence built entirely out of confidence and nothing else. We are aligned on that. It is on the roadmap. Let me circle back. None of those sentences contain information. All of them are doing the same job, which is buying time so the room does not notice that the person speaking does not know. I have done it myself, more than once, and I remember exactly how it felt each time, which was nothing like confidence and everything like holding my breath.

It has shifted again recently, in a way I find harder to sit with. It is no longer uncommon to be in a meeting and watch someone quietly type a question into a chatbot mid conversation, while the rest of the room waits in a silence that nobody quite names. I find it strange, possibly because I am older than the habit. If you book a meeting, I assume the people in the room came to share their own thinking with each other, not to outsource the thinking to a screen while everyone else watches. A fair number of meetings now seem to be less about what anyone in the room actually knows, and more about who can ask the model the sharpest question fastest. I am not sure that is a worse skill to have. I am fairly sure it is a different meeting than the one we agreed to have when we put it in the calendar.

What strikes me, looking back, is how much of leadership in tech gets built on top of that performance rather than underneath it. We reward the appearance of certainty far more often than we reward the harder, slower work of actually finding things out. A leader who says I do not know yet, give me until Thursday, often reads as weaker in the room than a leader who guesses with conviction and happens to be wrong. We have built whole cultures around rewarding the wrong half of that trade.

I carried that pressure for a long time before I noticed I was carrying it.

By the time I was leading the team I had worked alongside for three years, it had become something closer to an obligation. I read every procedure. I sat with other departments to understand how their work actually moved, not how the org chart said it moved. I gathered feedback from my own people and I actually listened to it, which is a smaller thing than it sounds and a bigger thing than most leaders manage. I pushed them hard. I gave them flexibility in return, time off when they needed it, early finishes when the work allowed it. That part felt fair. What never felt fair was the thing underneath it. The constant low hum of needing to be ready for any question that might land on my desk. Not because anyone demanded it of me. Because I had decided, somewhere back on the service desk, that this was what being trusted looked like.

Then I became a manager properly, and the maths stopped working.

You cannot know everything. Not because you are not capable, but because the surface area of the job grows faster than any one person’s memory can cover it. I had to learn to delegate, and I want to be precise about what that actually meant, because it is not the thing people assume. Delegating is not handing off a task because you are busy. It is asking, every single time, whether the person you are handing it to is being set up to succeed or quietly being set up to fail, and being honest with yourself about which one it actually is. I got that wrong more than once. I will not pretend I always called it correctly. But the discomfort of getting it wrong taught me more about leadership than any answer I ever had ready in a meeting.

There was a particular kind of fear underneath delegation that I do not think gets talked about enough. It is not really a fear that the work will be done badly. It is a fear that you will be exposed as someone who did not actually need to be the one holding all the answers in the first place. If someone else can solve the problem without you, what was all that knowing for. I think a lot of managers hoard answers without realising that is what they are doing, because letting them go feels uncomfortably close to admitting they were never as essential as the job title implied. I had to sit with that discomfort directly before I could let go of anything properly. It did not happen once and stay finished. I had to do it again every time the team changed, every time a new person joined who expected me to have the answer simply because of where I sat in the org chart.

What kept me afloat in those years was that I was always taking knowledge in. From people who would sit beside me and show me, often too quickly, often without patience, because they had their own ticket queue piling up. I learned to pay close attention in those moments because I knew I might not get a second explanation. From books. From the web, back when pages still loaded slowly enough that you had time to think about the question again while you waited.

And then, somewhere around 1998, a colleague said something that changed the shape of all of it. Go to Google. Just Google it.

I was on AltaVista at the time.

It worked, mostly. It was not elegant. Then Google arrived, and the first time I typed in an error message exactly as it appeared on screen and found someone, anywhere in the world, who had hit the same wall and written down how they got past it, something in me relaxed. Not because I no longer needed to know things. Because I realised I was not alone in not knowing them. Somebody out there had already paid the cost of figuring it out, and had chosen, for nothing, to leave the answer behind for the next person.

There is a generosity in that I think we have stopped noticing because it became so normal. An entire generation of engineers learned their craft not from manuals or courses but from strangers who had hit the same wall at two in the morning and decided, for nothing, to write down how they climbed over it. Forums full of people answering questions they had no obligation to answer. It was one of the strangest and best things the internet ever produced, and we absorbed it so completely that within a decade it had stopped feeling like generosity at all. It just felt like how information worked.

Leave a comment

For a stretch of years, into the mid 2000s, it was completely ordinary to be sitting in a meeting and hear someone say, hang on, let me Google it. And the room would wait, comfortably, while someone found the answer on a machine instead of in a colleague’s head.

I noticed the shift without naming it at the time. We were spending more time searching than asking. Not because asking had stopped working. Because searching was faster, and we were all under the same pressure to move quickly, and if the answer was already sitting out there for free, why would you spend someone else’s time getting it?

I do not think that decision was wrong, exactly. I think it was the first domino, and we did not see the rest of the row yet. What got quietly lost in that trade was not the answer. It was the conversation that used to happen on the way to the answer. The version where you turned to the person next to you and the act of explaining the problem out loud was half of how you solved it. You do not get that from a search box. You get the answer and skip the explaining entirely, and for a long time none of us thought that was a cost worth naming.

Even so, searching was still fundamentally an act of asking.

You typed the question yourself. You chose the words. You read several answers, weighed them against each other, and decided which one fit your actual situation, because the page in front of you was a list of other people’s attempts, not a single confident voice telling you the answer was settled. There was still a small act of judgment sitting in the middle of it, even if we stopped noticing it was there.

Then Covid arrived, and whatever was left of the habit of walking over to someone’s desk got removed entirely, for years, for some people permanently. I sat at home trying to get things done with the help of strangers’ generosity online, the same generosity that had started with that first AltaVista search, just compounded a thousand times over by then. There was real autonomy in it. I will not take that away from the period. But I missed the version of work where solving a problem meant standing next to another human being while you did it.

I tried to claw a little of that back with the teams I led. Virtual lunches. Coffee breaks that were allowed to be aimless. Social time that had nothing to do with delivery dates. It worked, I think, because we were all adults who already remembered what an office felt like, so we knew what we were trying to recreate. Remote work gave us a freedom none of us had been offered before. I will defend that part of it without hesitation.

What I noticed, slowly, over those years, was that the questions people brought to a video call were different to the ones they used to bring to a desk. On a screen, you arrive with the question already mostly formed, because the cost of opening a call feels higher than the cost of wandering over and thinking out loud on the way. Something about distance makes you arrive pre-edited. I do not think that is a worse way to ask a question. I think it is a narrower one. The half formed, still figuring it out version of a problem, the one you used to talk through on the walk between two desks, mostly stopped happening, because there was no walk left to have it on.

But I also remember sitting in my car one ordinary morning, stuck in the kind of traffic that used to be invisible to me because I had simply accepted it as the cost of having a job, and realising, properly realising, how much stress I had carried for decades without ever once putting it down on the table to look at.

Then came the models.

We built machines that could predict, with a kind of fluency that genuinely startled people who had spent their whole careers around computers, what word should come next. The frenzy that followed was enormous. Doom on one side, salvation on the other. Neither one fully arrived. What arrived instead, quietly, almost politely, was something else entirely. Tools built so that you would never again need to hold the answer yourself. Intelligence, metered out on demand, priced by the token, available the moment you needed it and gone the moment you closed the tab.

That small act of judgment I mentioned, the weighing of several answers against your own situation, mostly disappears here. A model gives you one voice, delivered with total confidence regardless of whether it deserves that confidence, and the format itself nudges you toward accepting the first answer rather than comparing several. I do not think this happened because anyone sat in a room and decided to remove judgment from the process. The business underneath these tools rewards speed and fluency, because speed and fluency are what keep you coming back to pay for the next query. Nobody designed it to make us less careful. It just turns out that careful is slower, and slower does not retain a subscriber the way fast does. That is the system working exactly as its incentives point it, with no one person to blame for the shape it has taken.

Share

I want to be honest about my own relationship with this, because the mirror only means something if you actually look in it. I use these tools. I use them to help me forecast, to map out possibilities I might not have reached for on my own, and even when the output is mundane, it is genuinely useful for stretching my imagination further than it would stretch alone. That part is real and I am not going to write an essay pretending otherwise.

I think the honest version of this is closer to medicine than most of us want to admit. You do not take ten paracetamol at once because one worked well for the last headache. You take the dose that is actually needed, you read the packet, you ask someone who knows more than you if you are not sure, and you stay alert to what mixing it with everything else in your day might do to you. Nobody thinks of that as restriction. We think of it as basic care for something powerful enough to do real harm if you stop paying attention to how much of it you are taking. I do not think we have built that same instinct yet for how much thinking we are handing over in a single day, and I include myself in that.

But I do not think we are using these tools the way we tell ourselves we are.

There is a difference between offloading a cognitive task because it frees you up for harder thinking, and offloading it because not doing the thinking yourself has simply become the path of least resistance. The first is a tool. The second, I think, is dependency wearing the tool’s clothes. And the second one does not announce itself. It feels exactly like convenience right up until the moment you notice you have stopped being able to tell the difference between the two.

I have sat with teams while they worked through exactly this tension, and it rarely resolves cleanly. Someone will say the model saved them four hours on a problem they would have solved themselves anyway, slower. Someone else will admit, quietly, that they no longer try the problem before reaching for the tool, which is a different thing entirely. Both of those people think they are describing the same behaviour. They are not. One of them is augmenting judgment they already have. The other is letting the judgment atrophy without noticing, because atrophy is slow and convenience is immediate, and we are creatures who respond to the immediate thing far more readily than the slow one.

The engineers I respect most right now are not the ones refusing the tools out of principle. They are the ones who can tell you, specifically, which parts of their thinking they have handed over and which parts they have deliberately kept for themselves, and why. That distinction takes effort to maintain. Most of us, myself included on certain days, let it blur because maintaining it is tiring and the tool does not ask you to maintain it. It just answers.

I keep returning to one question lately, and it will not leave me alone.

What happens if we get to a point where none of us actually knows the answer anymore? Not because we are incapable. Because we stopped needing to be the one who held it.

What does it mean when a software engineer can no longer explain why a repository is structured the way it is, because the structure was generated rather than understood. What does it mean when a project manager cannot tell you why a project needs to happen in a particular order, only that the plan said so. What does it mean when everything around you works, every system, every workflow, every dashboard, but the why of it has become a black box you have quietly agreed never to open.

I think about the engineers coming up now, the ones a few years into their careers, the way I once was on that service desk floor. A lot of what made me competent back then was not talent. It was repetition. I broke the same kind of thing enough times that I started to recognise its shape before it fully broke. If the tool gets there first every time, what does that engineer end up recognising. What pattern gets to live in their hands instead of in a model they cannot see inside. I do not know the answer to that yet. I am not sure anyone does. But I notice that most conversations about this stop at productivity, how much faster the junior engineer ships, and almost never reach the second question, which is what they are no longer building inside themselves while they ship it.

The same question sits inside leadership too, not just inside the codebase. A manager who has never personally wrestled with a hard tradeoff, because the model summarised the tradeoff for them in a clean paragraph, is going to have a thinner instinct for the next hard tradeoff that does not come with a clean paragraph attached. Judgment is built the slow way, through friction, through getting it wrong in a smaller room before you have to get it right in a bigger one. I worry we are quietly removing the smaller rooms.

There is a quieter cost too, one that sits between people rather than between a person and a machine. When the answer used to come from a colleague, something passed between you in the asking. A small debt, a small trust, a reason to remember each other’s names and check in later on whether the fix had held. When the answer comes from a model instead, that exchange simply does not happen. Nobody owes anybody anything. The problem gets solved and nothing relational gets built on top of it. I do not think that is a tragedy on its own, in any single instance. But I think it adds up, across thousands of small exchanges a year, into teams that know how to solve problems together far less often than they used to, because they have quietly stopped needing each other to do it.

Does any of this make you feel secure.

Does it make you feel in control. I do not think it does, even when we perform confidence about it in meetings.

And there is a harder question sitting underneath that one. If you no longer know the answer yourself, how do you know whether the answer you have been handed is even true. Verification itself becomes a chore, something we skip because we are tired, because we are behind schedule, because the thing on the screen looks plausible enough and plausible has quietly become our new bar for true.

It was on the way back from that same drive to the sea that I got the clearest possible proof of this, in a way I am slightly embarrassed to put in writing. There were signs along the road home, more than one, telling us the route ahead was closed. One of them said, plainly, do not rely on Google Maps. The app on the dashboard disagreed. It showed the road open, the route clear, the same calm blue line it always shows. We followed the blue line. A few kilometres later we hit the dead end exactly where the signs said it would be, had to reverse out, and found another way round that was not far off but was not the way we had planned either.

What stayed with me was not the wasted twenty minutes. It was that I watched a physical sign, in clear daylight, telling me one thing, and a screen telling me another, and some part of me trusted the screen more, right up until the tarmac itself ran out from underneath that trust. I got annoyed at myself afterwards, and at the app, which felt slightly unfair given how much data that company has presumably gathered from people like me just to get a road closure wrong. But underneath the irritation was something more uncomfortable. I had let a sign, the oldest and most basic form of information there is, lose an argument it should have won instantly, simply because the other source arrived on a screen and screens have quietly earned a kind of authority that paper and paint never did.

I think about this every time I watch someone, often someone sharp, someone I respect, read an AI generated summary of something important and nod along without checking a single underlying source. Not out of laziness exactly. Out of exhaustion, or out of the same quiet conditioning I caught myself in on that road. We are all carrying too much already, and verifying takes energy we have already spent on six other things before lunch. So we borrow trust instead of earning it ourselves, and we tell ourselves that borrowing trust is the same thing as having checked. It rarely is.

It feels, some days, like we are living by proxy.

Someone, or something, is handing us a version of reality. A version shaped by filters we cannot see, assembled by processes nobody asked us to approve, and most of us nod along because questioning it costs more energy than we have left in the day. I do not think this is a story about a villain sitting somewhere deciding to deceive us. Nobody designed this outcome on purpose. It is a thousand small, reasonable decisions, made under real pressure, by people including myself, that add up to something none of us would have chosen if we had seen the whole shape of it from the start.

I do not have a clean way to close that thought, and I am not going to manufacture one. The discomfort is the point.

I can already hear the objection, because I have made it myself on other days. Every generation says this about whatever made the previous generation’s skill obsolete. People said it about calculators. People said it about spreadsheets replacing the army of clerks who once did long division by hand for a living. The work moved up a level, the argument goes, and the people doing it moved up with it, and panic about the lost skill always looks foolish a decade later once the new skill has settled into place.

I think that argument is right often enough that it deserves real weight, and I do not want to wave it away just because it is inconvenient for the essay I am writing. The calculator did not make mathematicians less capable. It freed them to work on harder problems than long division. I am genuinely open to the possibility that this is exactly that story again, just bigger and faster and more uncomfortable to live through while it is happening.

But I do not think every shift is the same shift just because the shape of the argument repeats. The calculator replaced a mechanical task. It did not replace the asking of the mathematical question itself, only the grinding arithmetic underneath it. What concerns me about where we are now is not that a tool got faster at the grinding part. It is that the asking itself, the formation of the question, the judgment about which question was worth asking in the first place, is increasingly something we hand over too. That is a different layer of the work than the one calculators ever touched. I am not certain that difference matters as much as I think it does. I am genuinely not certain. But I am not willing to assume it does not matter either, simply because the comforting version of the story has worked out fine before.

That uncertainty is, honestly, where I have landed for now. Not with an answer. With a sharper question, and an intention to keep asking it of myself before I let anyone else answer it for me.

I think that is, in itself, a kind of practice worth naming.

Not refusing the tools. Not pretending the calculator argument is wrong just because it is uncomfortable. Just refusing to let the question close before I have actually sat inside it for a while. Most of the damage I have seen done by new technology over the years did not come from the technology itself. It came from how quickly we stopped asking what it was actually doing to us, because the asking itself felt like standing in the way of progress. I do not think slowing down here is standing in the way of anything. I think it is the last place left where the answer is still genuinely ours to give.

But I have been sitting with the original question long enough now that I think I have found one honest, partial answer to it. Not the whole thing. Just a thread of it.

When I think back over every version of myself in this story, the slip of paper on the service desk, the manager learning to delegate, the person typing an error message into AltaVista for the first time, none of them were actually valuable because of the answers they held. The answers changed constantly. AltaVista’s answers got replaced by Google’s. Google’s are getting replaced right now, by something faster and far more confident and not always right. If the answer was ever really the thing worth being, I would have nothing left, because none of my old answers have survived.

What survived is something underneath the answer.

I tested that line on the model before I wrote it down here. I asked it, flatly, whether I was right, that nothing survives but the noticing. It did not flatter me. It pushed back, said I was romanticising the parts of the job that were already disappearing, that discernment without an answer underneath it is just a slower kind of guessing. I sat with that longer than I expected to. I still think it is wrong. But I noticed I had to actually think to disagree with it, instead of just nodding the way I nod at most things on a screen. That, at least, felt like the old way of finding an answer. The kind where you argued with someone next to you on the way there.

It is the willingness to stand in front of someone who needs help and say, I do not know yet, give me a minute, while actually meaning it instead of performing it. That sentence costs something to say honestly. It is far easier to guess with conviction. I think the people who earned real trust from me over the years, in any role, were almost never the ones who had the most answers ready. They were the ones who were honest the moment they ran out of answers, and then went and found the missing piece anyway, in front of me, without pretending the gap had never existed.

It is the part of me that learned, on a packed call centre floor with phones dead and a manager shouting, which question to ask first and which one could wait. Nobody taught me that directly. It came from doing it badly a few times and feeling the cost of getting the order wrong. That instinct does not live in a model. A model will answer whichever question you give it with equal confidence regardless of whether it was the right one to ask first. The choosing of the question is still entirely ours, and I think it always will be, because choosing requires caring about the actual outcome in a way a prediction engine has no stake in.

It is the years spent listening to a team closely enough to know what feedback meant underneath the words they actually used. Someone tells you the workload is fine and you can hear, if you have spent enough time with them, that fine is doing a lot of quiet work in that sentence. None of that shows up in a transcript a model could summarise accurately. It shows up in tone, in timing, in the half second pause before someone answers. You only catch it if you were actually paying attention to the person and not just extracting the information you needed from them.

I think who you are without the answer is whoever was doing the noticing the whole time the answer was sitting in your hands. The judgment about which question actually mattered. The discomfort you were willing to sit inside instead of immediately resolving. The trust someone placed in you not because you always knew, but because you never pretended to know when you didn’t, and you went and found out anyway, in front of them, together.

That part, I do not think a model can take from us. Only we can give it away, and I think a fair number of us are giving it away right now without noticing, the same quiet way we once stopped walking over to a colleague’s desk because the search box was simply faster.

I am not writing this to tell you to put the tools down.

I use them too, and I am not going to pretend that chapter of my own working life is over. I am writing it because I think the question underneath the convenience is worth asking out loud, on purpose, before the habit fully sets, rather than after.

I think that is part of why I wanted the sea that day, more than I realised while I was still driving toward it. Not for an answer. The sea has never once told me anything. I think I wanted to stand in front of something old enough and large enough that it had no opinion about automation, no answer to sell me, nothing metered, nothing priced by the token. Just the sound of it, doing the same thing it has always done, asking nothing of me in return.

I did not get that on the way home either, as it turned out.

I got a blue line on a screen telling me a road was open that a painted sign, standing right there in front of me, was telling me was closed. I believed the line. I am still slightly unsettled by how easily I did. I do not think I came back from that drive with a conclusion. I think I came back with the same question the sea never answered, just pointed now at something closer to home, sitting on a dashboard instead of a horizon.

If you read this far and felt something land, the simplest thing I can offer is this. Every Tuesday I send one essay like this one, free, straight to your inbox, no catch attached to it.

If that is useful to you, that is genuinely enough, and I mean that.

Maybe that is the whole report from this drive. Not a verdict on AI. A smaller thing. The blue line did not create my willingness to trust it over a hand painted sign standing right there in front of me. It just showed me how far that habit had already travelled, quietly, for years, long before any model existed to take the blame for it.

If this question matters to you too, subscribe. Every Tuesday I write about leadership, AI, and the quiet ways technology changes how we think, work, and lead.

About the Author
Diamantino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

And the endless discussion of what AI is, what it is capable of, the doomers, the utopians.

The coverage that followed had the same shape it always does. What does this mean for your workflow? What does this mean for your team? How do you use it effectively? How do you stay ahead of the people who are also figuring out how to use it?

I read through most of it. And then I closed the tab and sat for a while with a feeling of emptiness, knowing that all that information was somewhere in a documentation section of a particular site. But none of it asked the question that mattered most.

What is all of this doing to us?

Not to our output. Not to our productivity or our competitive position or our quarterly numbers. To us. To the people sitting at the keyboards. To the engineers and the leaders and the teams who are now, whether they chose it consciously or not, reorganising their working lives around a tool that is getting smarter at a rate that human beings are not.

I have been in technology long enough to recognise the pattern. Every release is framed as a gain. Every announcement is a celebration of what has become newly possible. And the energy around that is real. I feel it myself. There is something genuinely compelling about a tool that keeps expanding what it can do.

But there is a question that does not get asked at the launch party. And I think the cost of not asking it is starting to accumulate in ways we are only beginning to see.

Leadership as a verb is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

The conversation we keep skipping

I want to be precise about what I am pointing at, because it is easy to hear this kind of concern and file it under the wrong category.

I am not talking primarily about job displacement, though that is a real and serious issue that deserves its own honest conversation. I am not talking about hallucinations or errors or safety benchmarks, though those matter enormously and are not yet solved. I am talking about something that operates at a slower frequency and is harder to see because it does not produce a visible incident. It produces a drift.

When a model gets better at reasoning, something changes in the room where humans used to reason together. The room gets quieter. The pauses get shorter. The arguments that used to happen because two people had genuinely different intuitions about a problem start to happen less, because there is now a third voice that is faster than both of them and carries no emotional stake in being right.

When a model gets better at writing, something shifts in the relationship between a person and the effort of forming their own thoughts. The blank page, which used to be a necessary friction, a space where you had to hold the discomfort of not yet knowing what you meant long enough to find out, becomes optional. You can skip it. You can start with output instead of starting with silence.

When a model gets better at coding, something moves in the space where a junior engineer used to spend three uncomfortable hours on a problem they could not crack. Those three hours were not inefficiency. They were formation. They were the process by which someone became a person who could hold a complex system in their mind and navigate it. When you remove that struggle, you do not just save time. You interrupt something that was quietly building a capability that cannot be installed any other way.

I have watched this happen in real teams. In engineers I have worked alongside and coached. In myself, on the days when I reach for the tool before I have even sat with the problem long enough to know what kind of problem it actually is.

The output gets faster. The thinking gets thinner. And because the output is visible and the thinking is not, the thinning does not show up anywhere that anyone is measuring.

What these systems are, and what they are not

Part of what makes this conversation difficult is the language we use to describe what is happening. We call these systems intelligent. We talk about them reasoning, understanding, and knowing things. We have absorbed, without quite deciding to, a framing that places these tools somewhere on a continuum with human cognition, just faster and more capable in certain domains.

I noticed it in myself last week, mid-meeting, reaching for the laptop before I had even finished hearing the question. Not because the question was hard. Because some quiet calculation had already run: the tool will be faster than I will. That calculation is the framing doing its work, and it does its work in seconds, before anyone in the room has time to object.

That framing is doing significant damage, and not because it is purely wrong. It is wrong in the ways that matter most.

A large language model is trained on human text. It becomes extraordinarily good at the patterns of how human beings write and think and argue. It can recombine those patterns in ways that produce outputs that look like reasoning and read like insight. In a narrow, functional sense, something remarkable is happening inside these systems.

What they cannot do is understand. Not in the way a person understands something. Not with a body that has carried grief or made a promise they later had to break. Not with a history of being wrong about something that cost them a relationship and having to live inside that for years. Not with the capacity to sit across from another human being and feel, in the texture of the silence before they speak, that something underneath their words is different from the words themselves.

That gap is not a version problem. It is not something that the next release will close. It is a different category of thing entirely. And when we describe these systems as if that gap were merely a technical limitation being worked on, we do something subtle and consequential. We begin to calibrate human intelligence against machine output and find the human version slower, messier, and harder to justify.

That recalibration is not neutral. It does not stay contained to how we think about productivity. It starts to shape how people feel about their own minds. About the value of the slow, uncertain, embodied process of thinking something through without help. And once that feeling takes hold in a team, it changes what the team is willing to do.

How dependency builds without anyone deciding to build it

I want to describe a pattern I have observed carefully, because it is easy to dismiss as anecdote until you have seen it enough times that it starts to look structural.

An engineer begins using an AI assistant. The initial relationship is uncomplicated. They reach for it on the repetitive work, the boilerplate, the tasks that feel like friction between them and the problem they actually want to solve. This is a sensible choice. There is no reason to write the same kind of code manually that you have written a hundred times before. The tool handles it. The engineer moves on to the interesting part.

But six months later, something has shifted. They are reaching for the tool earlier in the process. Not just on the repetitive parts, but on the design questions. On the architectural decisions. On the places where the real thinking is supposed to happen. They are not doing this because they cannot think without the tool. They are doing it because the gap between what the tool produces in thirty seconds and what they would produce after twenty minutes of genuine effort has started to feel like evidence of their own inadequacy. The comparison is unfair, but that does not make it less real as an experience.

The tool did not engineer this feeling deliberately. There is no malice in a model. But the system behind the tool, the company that built it, the business model that depends on it becoming indispensable, the incentive structure that rewards increasing usage and would be damaged by decreasing it, that system understands exactly what it is doing. It is not designing for augmentation. It is designing for reliance. Those are different things, and the difference matters enormously.

When you build something that gets measurably smarter every year while the person using it stays roughly the same, the trajectory of that relationship only goes in one direction. You do not end up with someone who has been enhanced. You end up with someone who has been made dependent in ways they may not fully recognise, because the dependency formed gradually and because each individual step along the way felt entirely reasonable.

This is not a side effect that the industry is working to eliminate. In many cases, it is the product.

What happens to teams that stop disagreeing

There is a social and institutional dimension to this that gets even less attention than the individual one, and it may be more consequential at scale.

Teams think together. Not just in the sense that individuals who happen to work near each other each do their own thinking, but in the deeper sense that the quality of collective reasoning in a well-functioning team is genuinely higher than the sum of its parts. That happens through argument. Through the friction of people with different instincts pushing against each other until something that neither of them could have reached alone becomes visible.

That process is slow. It is uncomfortable. It requires people to hold uncertainty without resolving it prematurely. It requires a tolerance for being wrong in front of your colleagues. It requires the kind of psychological safety that takes a long time to build and very little time to damage. It is, in other words, exactly the kind of thing that a fast, confident, always-available tool makes it tempting to skip.

When a team starts routing its hard questions through a model before it routes them through each other, several things happen. The arguments get shorter, because there is already an answer on the table and arguing with an answer is harder than arguing with a colleague who is still forming their view. The junior voices get quieter, because the model does not hesitate the way a junior person hesitates, and hesitation is often where the most genuine thinking lives. The range of perspectives that gets considered narrows, because the model draws on what has already been written and published, not on the particular context and history and local knowledge that lives only in that room.

Over time, the team gets more efficient and less wise. And because wisdom is not a metric that anyone is tracking, the loss does not register until something goes wrong in a way that better collective thinking would have prevented.

Organisations that normalise machine-first thinking do not just change their workflows. They change the culture of how people relate to uncertainty and to each other. That is a long-term cost that will be very difficult to reverse once it is established.

When a team stops disagreeing, it does not become more aligned. It becomes more brittle. The arguments that felt inefficient were doing something the efficiency could not replace.

Who benefits, and why they will not say so

Every time a new model is released, the question asked loudest is what it can do. The question almost never asked is who benefits from you not being able to do without it.

Augmentation would mean building something that makes you more capable and then stepping back. What is actually being built is something that makes itself progressively harder to remove from your workflow, that captures your data and your patterns in ways that make switching costly, that is priced to create habitual use and structured to reward increasing dependence.

What makes it effective is that the dependency does not feel like dependency while it is forming. It feels like getting better at your job. Only later, when you try to think through a hard problem without it and feel the absence more acutely than you expected, do you begin to understand what has quietly happened.

The question for leaders is not whether to engage with these tools. That question is already answered. The question is whether you understand the power dynamics you are operating inside, and whether you are making deliberate choices about which parts of your team’s capability you are willing to trade and which parts you are going to protect.

There is no benchmark for what happens to the engineer who stops working through problems by hand and loses, over two years, the ability to hold a complex system in their head without external scaffolding. There is no metric for the leader who has outsourced the uncomfortable ambiguity of difficult decisions to a prompt and slowly lost the tolerance for sitting with uncertainty that good leadership requires. The industry measures what AI can do. Nobody with any institutional power is measuring what it does to us. And that asymmetry is not accidental. The companies releasing these models do not benefit from a public conversation about the cognitive and social costs of their products.

I use these tools. I used one while preparing this essay, and I am saying that directly because being transparent about the tool is itself a form of intellectual honesty that I think matters here. The thinking is mine. So is the uncertainty that came with it. The tool helped me find what I was trying to say. It could not feel why it needed to be said, or what was at stake in saying it clearly.

What this looks like this week

This does not require a policy or a tool ban. It requires noticing, once, somewhere small. Pick one:

• In your next team meeting: before anyone opens a model, ask “what do we think, before we check.” Let the disagreement happen first.

• Before you reach for the tool yourself: sit with the problem for as long as it takes to form one rough guess of your own. Then check it against the machine, not the other way round.

• With one junior person this week: hand back a question they tried to outsource. Not as a test. As a kept door.

None of these cost you the tool. They just decide, on purpose, where the human goes first.

What staying human actually requires

I am not making an argument against using these systems. That would be dishonest, given that I use them, and it would also be the wrong frame. The question is not whether to use them. The question is whether you are using them with clear eyes about what you are trading and what you are protecting.

What staying human requires, in this particular moment, is something harder than refusing a tool. It requires naming the question out loud, in the rooms where it is most inconvenient to ask.

What is this doing to us?

Not as a philosophical exercise. Not as a rhetorical gesture toward responsibility. As a real leadership question with real stakes and a real obligation to sit with the discomfort of not having a clean answer.

What skills are we in danger of losing if we keep handing over the work that built them? What judgment gets weaker when we stop exercising it? What becomes irreversible if the reliance runs deep enough for long enough, not just in individual people but in the culture of entire organisations?

And then the harder question. What do we choose to protect, and what are we willing to do to protect it?

Not by refusing the tool. By being deliberate about where the human has to stay in the loop, not because it is more efficient, but because some capabilities only form through the doing of them. Because struggle is not waste. Because the engineer who spends three hours on a problem they could have handed over in thirty seconds comes out of those three hours with something the model does not have and cannot give: the knowledge of what it feels like to be stuck, and the earned confidence that they can find their way through.

There is a kind of knowledge that cannot be downloaded or transferred. It is built in the body and the will, through repetition and failure and the specific sensation of something finally making sense after a long period of it refusing to. That knowledge is what we are here to protect. Not as a nostalgic gesture toward the way things used to be done. As a recognition that what makes human judgment irreplaceable is precisely the difficulty of building it, and that difficulty is not a problem to be solved but a process to be respected.

The releases will keep coming. The benchmarks will keep climbing. The capabilities will keep expanding.

But the question that does not get asked at the launch party remains the most important one.

Not what can we do with it.

What will it do to us.

And what are we going to do about that.

About the Author
Diamantino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

A fellow human being.

The headline was clean, confident, and brutal, You can’t beat AI. AI is smarter and cheaper every year. Do you?

I read it. Then I read it again.

It was not the argument that stayed with me. I have read versions of that argument before. What stayed was the feeling underneath it, something between discomfort and a quiet kind of grief. The kind of feeling you get when someone says something in a way that makes you feel smaller than you were before you read it.

It felt cold.

It also felt strange to see people commenting in appreciation without noticing the deeper cost of the frame itself.

People are being conditioned, and that conditioning is profitable for someone. I thought.

What the post was saying

The post made a practical argument. AI lowers the cost of average work. Average output is now cheaper than it has ever been. Therefore, if you want to remain economically relevant, you need to stop selling hours and start selling taste, judgment, and the capacity to make final calls.

That is not wrong. In a narrow labour-market sense, it maps to something real. The economics of text, analysis, basic code, and routine cognitive work have shifted.

But there is a difference between a true observation and a complete one.

And there is a difference between naming a market reality and choosing what to do with that naming.

This post chose a specific frame, competition. It treated human beings and AI as if they were in the same race. As if the only question worth asking were, what can a person still do that software cannot do yet, cheaply enough to justify the price?

That frame is not neutral. It carries a philosophy.

The units problem

Here is what I think the post did, even if it did not mean to.

It treated people as units of productivity.

Not as people. Not as beings with histories, relationships, moral weight, and the particular irreducible quality of having lived a specific life. As units of productivity. Variables in an equation. Line items whose cost is justified only by output.

This is not a new frame. It has been the dominant frame of industrial capitalism for centuries. What is new is that it is now being applied to cognition itself.

We have spent a long time treating physical labour as fungible. Now we are extending that logic to thought.

The argument becomes, if a machine can think, and thinking was your last advantage, then you are either differentiated or you are a commodity.

But the question is whether the thing we call AI is actually thinking in the way humans think. AI is a field of study. What most people are using today are large language models, statistical systems based on transformer architectures, trained on vast amounts of text. They are powerful pattern-completion engines, not conscious beings.

They do not understand. They do not care. They do not have lived experience. They do not hold responsibility. They generate outputs that can look intelligent because they are fluent, not because they are aware.

There is a word for what that kind of framing does to people, dehumanisation.

Not in the dramatic sense. Not in the sense of cruelty or malice. In the clinical sense. It removes human context from a human being and reduces them to function.

I do not think the author woke up that morning trying to make anyone feel less than human. But intention does not determine impact.

And the impact of many posts with this structure, arriving in feed after feed, is cumulative. It builds a climate where people are trained to ask not “what kind of life do I want?” or “what does my work mean to the people it touches?” but “how do I stay economically superior to a language model?”

That is a very small question. And it is becoming the question.

What AI actually is

Before we talk about beating something, it helps to be precise about what we are being asked to beat.

There is no singular AI. That word is doing a lot of marketing work. It covers everything from spam filters to computer vision to robotics to large language models, with very little in common beyond computation and commercial usefulness.

What most people mean in the productivity conversation is simpler, large language models.

These are statistical systems trained on enormous quantities of text. They are exceptionally good at pattern completion. They produce fluent language. They can summarise, reformat, generate first drafts, and approximate reasoning in domains where the answer often resembles the average of many previous answers.

They do not verify their own outputs. They hallucinate with confidence. They have no continuous memory. They do not understand consequences in the world. They cannot be held accountable. They do not grieve. They do not learn from experience the way a person learns from experience. They have no stake in what happens next.

These are not minor limitations.

Share

Some are structural. Some are built into how these systems work.

So when we ask whether humans can “beat AI,” we are often comparing a person to a tool that is being described far too broadly and far too vaguely.

That comparison is not honest. And it is not useful.

A human being and an LLM are not the same kind of thing. One has lived experience, accountability, and moral weight. The other produces language at scale.

The business model behind the message

Here is the question the original post did not ask,

Who benefits from you believing you are in competition with AI?

The answer is not mysterious.

Every company selling AI tools benefits. Every platform that thrives on engagement benefits. Every course, conference, and certification business built around AI anxiety benefits. The venture ecosystem that has bet heavily on this moment benefits from making the shift feel total, inevitable, and urgent.

That does not mean the tools are fake. It does not mean the change is not real.

It means there are powerful incentives to make this moment feel more like extinction than it may actually be.

Fear sells better than nuance. Urgency sells courses. Obsolescence sells subscriptions.

So when a piece of content arrives saying you cannot beat this thing, it is worth asking what it is trying to get you to do.

Usually the answer is simple, adopt something, buy something, move faster, panic productively.

That is not a conspiracy, it is how attention economics work.

Naming that is not cynicism. It is literacy.

What gets lost

A productivity-only frame leaves out too much.

It leaves out care.

Care as labour. The nurse who comes back because something felt off. The teacher who notices a student has gone quiet. The manager who holds something difficult in confidence because they promised they would.

LLMs can mimic the language of care. They cannot care.

They can generate the shape of empathy. They do not bear the weight of it.

When we reduce human value to competitive cognitive advantage, care becomes invisible because it does not always show up in a metric. It becomes the inefficiency that productivity thinking teaches us to eliminate.

It also leaves out accountability.

When something goes wrong, who is responsible?

The vendor says it was the user’s choice. The user says they followed the system’s recommendation. The system says nothing. The person harmed has no one to face.

This is already happening in hiring, medical decisions, content moderation, credit systems, and predictive policing.

Human beings are accountable in a way systems are not. Not because we are always right. We are not. But because we can be faced. We can be asked to explain. We can feel the weight of what we have done.

That capacity is not a weakness. It is one of the foundations of trust.

And it leaves out lived experience.

Something happens in a life that no training data can replicate. The specific texture of having been through something. A restructuring. A dismissal. A hard conversation. A mistake that changed the way you listen the next time.

That is not in the model. It cannot be.

It is not text. It is history made physical.

The question the post forgot to ask

The post asked, how do you stay economically relevant in an age of AI?

That is a real question. Many people are afraid, and the fear is not irrational.

But there is a prior question,

What is technology for?

If the answer is to make production cheaper, then the logic of the post follows. AI makes cognitive work cheaper. Find a way to differentiate or accept the price.

But if the answer is to expand human possibility, reduce drudgery, and create more room for judgment, care, and presence, then the frame changes completely.

Then the question is not how to stay economically superior to a language model.

The question is what we do with the time and attention that get freed up.

What kinds of human work do we want to protect?

How do we make sure organisations use these tools to deepen human value instead of flattening it?

Who bears the cost of transition?

These are not naive questions. They are the real ones.

A more humane frame

I do not think the original post was malicious. I think it was trying to be useful.

And in a narrow sense, it was.

It pointed to a real shift. It reminded people that average work is becoming cheaper. It told them to focus on judgment, taste, and decision-making.

But a more humane version would say something different.

It would say, AI should reduce drudgery so that the things that require full human presence get more of our time and attention.

It would say, the organisations that serve people well over the next decade will not be the ones that replaced the most humans with the most tools. They will be the ones that used tools to deepen the quality of human work.

It would say, your value is not conditional on being faster than a language model.

Your value is that you are a person. In relationship with other persons, carrying a history, capable of accountability, of care. Of being changed by experience.

That is not sentimental.

It is the foundation of any society worth defending.

This is not an instruction to ignore the shift. The shift is real.

But there is a difference between adapting to a change and accepting the philosophy that comes packaged with it.

You can use these tools and still refuse to reduce your team to a list of functions that have or have not been automated yet. You can be honest about what AI does well and still as, what kind of organisation do we want to be when this settles? What do we owe the people whose work is changing?

Those are not soft questions. They are the ones that will determine what is left standing after the efficiency gains have been captured.

The leaders who ask them now will not necessarily slow anything down.

They will know what they are building and why.

What the feeling was telling me

The feeling at the beginning of this was telling me something. Not just discomfort. Recognition.

I have seen this pattern before.

During the crypto frenzy, the message was consistent: this is inevitable, the early movers win, the people who wait will be left behind. The urgency was engineered. The FOMO was a product. Millions of people boarded that train. Some made money. Many lost savings they could not recover. The people who designed the on-ramp did not.

The phone scammer works the same way. Your account has been compromised. Act now. Do not hang up, do not check with anyone, do not think. The entire mechanism depends on bypassing the pause.

What I noticed in the post I read, and in the dozens like it, is the same structure. The urgency is pre-loaded. The conclusion is the premise. You are already behind. You are already the expensive option. The only question left is whether you will accept that fast enough to do something about it.

The people who benefit from that message moving fast are not the people reading it.

That is not cynicism. It is pattern recognition.

And sometimes the most important thing a leader can do is pause long enough to notice the pattern.

We need to pay attention not only to what is being said, but also to who is saying it.

About the Author
Tino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

That feeling is the whole essay. Not the technology. Not the steps. The feeling. The moment you realise something you built is working without you watching it. That your intention, clearly expressed, became action you did not have to take yourself.

This is what I want to talk about. Not how to build an AI agent. Not a tutorial with steps and screenshots. Something harder. The mindset you need before you write a single line of anything. The understanding that separates people who build systems that work from people who build systems that impress no one, including themselves.

Because most people are getting this wrong. And most companies already know it.

This has always been about control

Before you start, understand what you are actually after.

Organising things, planning things, building systems that run predictably. None of this is really about efficiency. It is about control. The assurance that tomorrow will resemble today in the ways that matter. That the bills will be paid. That the work will get done. That the life you are building will hold its shape.

There is a reason so much of Western culture suffers from stress. And there is a reason the people trying to decompress keep being told to live in the moment, to accept chaos, to surrender to what will be. Que sera, sera. For most of us, this is almost an impossibility. You cannot relax into an unseen future when the mortgage, the bills, the life you are holding together, all depend on decisions with causes and consequences. The anxiety is not irrational. It is the correct response to real uncertainty.

We automate for the same reason we make lists and set alarms and build routines. To hold the uncertainty back a little. To create small pockets of predictability in a world that does not offer them by default.

I think about what Frederick Winslow Taylor did to work. The factory model. Every motion measured, every task broken into its smallest repeatable unit, every worker optimised for output. It gave us productivity and it gave us alienation in equal measure. And I look at practices like DevOps, Agile, the whole vocabulary of modern organisational efficiency, and I see the same logic in softer clothes. The factory did not disappear. It learned to use words like flow and iteration.

These things we call AI agents are the newest expression of that same impulse. Control, at scale, dressed up as intelligence. And there is a particular kind of person, usually an engineer, who will tell you proudly that they no longer write code themselves. That their hundred agents are busy doing things, all managed from a phone. I find that image genuinely funny. And also worth examining.

Share

Someone somewhere said we would sell intelligence by the metre. A commodity. Like water, like electricity, like the oil before both. And I think they were right about the direction, even if the destination is still unclear.

I could have written this essay as a tutorial. How to use any of the popular tools. That would have been easier to write and probably easier to find. But I refuse to join the chorus of “just use it or lose it.” The pressure to adopt without understanding. The anxiety that if you pause to think you will fall behind. I struggle with this myself. Genuinely. Because if I stop thinking carefully about things, if I outsource the reasoning as well as the execution, I am not sure what I am left with. The organ in my skull that helps me navigate this world is not something I want to hand to something that, at this point, cannot comprehend reality. And yet that same thing is now shaping decisions that are ending some people’s lives. Not metaphorically.

And it is worth saying plainly: this conversation is happening mostly in the West, among people who can afford twenty dollars a month, or two hundred. But when I speak with someone from Nigeria, Mumbai, Brazil, I notice something different. They squeeze every drop. They think carefully before passing anything to an AI, because they cannot afford to waste the exchange. There is no superfluous prompt. No casual experiment. The constraint sharpens them. It is a remarkable thing to witness, and an uncomfortable one, because it shows that the creativity this technology promises is already unevenly distributed before a single agent is built. Gen Z, meanwhile, has grown up assuming the web is just there, like oxygen, which makes the dependency harder to see and harder to question.

This is the same behaviour companies have normalised on a larger scale. Move fast. Adopt early. Work out the value later. And when the platform goes dark, when the credits run out, when the service changes its terms, the dependency is already built in. I know people building entire products on credits they do not own. I see competitors sharing cloud infrastructure with the very companies they are competing against. It is a strange tangle when you look at it all at once.

It is here to stay. How it unfolds, I believe, depends on whether ordinary people decide they have a voice in it. Not just technologists. Not just investors. Everyone who lives inside the systems being built.

What you actually need to understand to get the best from an automation agent

Given all of that, what do you actually need to understand to get the best from an automation agent?

Start here: they are still flawed. Mistakes will happen. The question is whether you have built something that makes mistakes visible and correctable, or something that makes them invisible and compounding. Everything that follows is an answer to that question.

The word “agent” is getting in the way

Let me remove a word that is causing confusion.

When people hear “AI agent” they imagine something sentient. Something making decisions with intention. Something that understands you. Something, if we are honest, that might care about getting it right.

It does not.

What you are building, in almost every case, is an automation agent. A system that takes instructions written in natural language and executes them against real tools, real data, real consequences. The “AI” part is the interpreter. The part that reads your instruction in plain English and translates it into action. But the action itself is deterministic. It follows a path. Your path.

Understanding this distinction changes everything about how you build.

Here is the human problem underneath it. We anthropomorphise relentlessly. We name our cars. We thank the GPS when it finds us a parking space. We apologise to the chair we walked into. We have always done this. It is not stupidity. It is how human perception works. We are pattern‑matching animals who find faces in wood grain and intention in weather.

The risk is not that you find your agent charming. The risk is that you extend it the kind of trust you extend to a person. That you stop auditing its output because it has been reliable lately. That you assume it understood what you meant rather than checking what it did. An automation system you trust like a colleague is an automation system that is one edge case away from a problem you did not plan for.

Keep a useful distance. Not cold. Not suspicious. Just clear‑eyed. You are the architect. The system is the scaffolding. The clay and the chisel. You are the one who knows what the shape should be.

Before you build anything, understand what you are automating

The most common mistake I see is people who start with the technology.

They install the tool. They connect the API. They write a prompt. They run it. Something happens. It is not quite right. They tweak the prompt. Run it again. Something else happens. They are now two hours in, iterating on something they never properly defined.

This is not building. This is tinkering. Tinkering has its place. But tinkering with no map produces nothing you can rely on.

Before you build, you need to understand the process you are automating with the same depth you would use to explain it to a new hire on their first day. Not the high‑level version. The actual version. Every edge case. Every exception. Every point where a human would normally make a judgment call without naming it as a judgment call.

Ask yourself this: if I wrote down every step of this process on a piece of paper, would a careful twelve‑year‑old be able to follow it without asking me a single question?

If the answer is no, you are not ready to automate it.

This is not about making things complicated. It is about making the invisible visible. Most processes we run daily are invisible to us precisely because we have done them so many times. Our hands know the steps. Our judgment fills the gaps without being asked. Automating an invisible process is how you get a system that works perfectly in normal conditions and fails silently the moment something unusual happens.

Why are you actually automating this?

This is the question most people never ask. And it is the one that matters most.

The honest answer, for many people, is this: because it feels like the right thing to do right now. Because everyone is talking about agents. Because the tools exist. Because it seems faster.

That is not a reason. That is a feeling. And building a system around a feeling produces a system that solves a problem you may not actually have.

Automating a broken process does not fix the process. It accelerates the breaking. You get the same wrong output, faster, at scale, with less visibility into where it went wrong. That is not progress. That is expensive noise.

Before you write a single instruction, you need to be able to answer one question in two sentences or fewer: what specifically changes for a specific person if this works? Not in theory. Concretely. What do they stop doing. What do they start doing instead. What does that free them for.

If you cannot answer that, stop. Do not build yet. Watch the process run three more times with human eyes. Talk to the person doing it. The real answer is almost always in the gap between the process as documented and the process as actually lived. That gap is what you are automating. Not the steps. The gap.

Prerequisites, requirements, and the trial that actually teaches you

There is a framework I return to every single time I build something new. Not because it is elegant. Because it has saved me from myself more times than I can count.

Three questions before anything else.

1. What does this need to have access to? These are your prerequisites. Data sources, credentials, file paths, external services, permissions. Write them down. Every single one. Because you cannot build a system around a dependency you have not named. And you cannot test a system against a dependency you discovered halfway through.

2. What must this system do, and how must it behave? These are your functional and non‑functional requirements. Functional is what it does: produces a summary, sends an email, updates a record. Non‑functional is how it does it: responds within a defined time, never contacts someone without explicit confirmation, logs every action so you can audit it later. The non‑functional requirements are where most people get caught. Easy to forget until the system does something you never told it not to do.

Then you trial.

Not the full system. A small piece. The most uncertain piece. The part you think you understand but cannot prove yet. You run it with real data. You watch what it does. You do not assume. You observe.

Trialling is not a step in the process. It is the philosophy of the process. The assumption that what you think will happen and what actually happens are different things, and the only way to close that gap is to create the conditions where reality can surprise you.

Most people rush this stage because they want to see the finished thing. The finished thing is the wrong goal. The goal is a system you trust. Trust is earned through iteration against reality, not through theoretical correctness.

One more thing about trialling. Modern agents can persist. They can keep going, trying different approaches, circling back, until they arrive at something that looks like a solution. This is now genuinely possible. It is also one of the more misleading features if you do not understand what it means.

A plausible solution and the right solution are not the same thing. An agent that finishes is not an agent that finished correctly. Before you run anything, define what correct looks like. Write it down. Because a system with no definition of done will produce a definition of its own. And you may not recognise it when it arrives.

Tokenisation is not a technical detail. It is a budget.

Here is the thing nobody explains when they talk about agents at scale.

Every instruction you give your system, every piece of history it carries forward, every document you ask it to hold in memory while it works, has a cost. Not just in money. In reliability. In the probability that your system reaches the end of the task without pausing, degrading, or stopping in a state you did not plan for.

Think of it like a meeting room. You can invite as many people as the space allows. But the more people in the room, the harder it becomes to make a sharp decision. Too many voices, too much context, and the output blurs. The model works the same way. Push too much into its context and the precision softens. Hit the limits of your plan and the task stops mid‑run. You come back to find it at step four of twelve. The work is incomplete. The data is in an intermediate state. Whatever it touched is now partly done.

I have seen well‑designed systems fall over the first time someone ran them on real data at real volume. The demo worked. Production did not. The difference was almost always the same thing. Too much history carried forward. Too much documentation included because it might be relevant. Too little discipline about what actually needed to be in the room.

The practice is simple in principle and requires constant attention in execution. Include only what the model needs to do the current step. No more. A focused, scoped instruction set is not a limitation. It is the architecture of a system that actually finishes what it starts.

Your local setup is your independence

There is a version of AI agent infrastructure where everything lives in someone else’s cloud. Their model, their servers, their pricing, their uptime, their terms of service.

That version is a rental agreement. And like any rental agreement, the terms can change.

I want to argue for ownership. Or at least for an architecture that behaves like ownership.

Running models locally is more possible than most people realise. You can run capable open‑source models on hardware you already own. Not the most powerful models available, but powerful enough for many real tasks. Summarisation. Classification. Extraction. Conversation with your own documents. A home server or a reasonable laptop is enough to begin.

But the larger argument is not about which model you run. It is about your knowledge base.

Your knowledge base is the information you have curated, organised, and made available to your agents. Your notes. Your processes. Your reference material. Your institutional memory. It is what transforms a general‑purpose model into a system that knows your specific context. That speaks your language. That understands your constraints without being told them each time.

Most people build their knowledge base inside a tool. Inside a platform. Inside someone else’s structure. And when that platform changes its pricing, or shuts down a feature, or gets acquired, the knowledge stays but the coherence goes. You have to rebuild. Or you lose it quietly over time.

Here is the design principle that changes this. The knowledge base is the soul of your setup. The model is interchangeable.

If you build your knowledge base in open, portable formats—plain text, Markdown, structured documents you own—you can swap the underlying model whenever a better one arrives. You can move from a cloud provider to a local model. You can upgrade without starting over. Your agents will still behave consistently because the knowledge that shapes them travels with you.

Back it up. Version it. Treat it the way you would treat the original manuscript of something you could not rewrite from memory. Because you could not. That accumulation, the way you structured it, the connections you drew between pieces, the decisions you recorded, that is yours. No model produces that. You did.

The possibilities you cannot see yet

I want to give you two pictures. Not to impress you. To show you that the ceiling is higher than you think.

The first is a house. Lights that adjust when the sun moves. Heating that learns the difference between a working day and a slow morning. A front door that knows your rhythm better than you do. A voice that can tell you what is in the fridge, book a table, set a reminder, send a message, and do it all from a single sentence spoken out loud. None of this requires expensive hardware. None of it requires a developer. It requires someone willing to understand what they want, define it clearly, and build it in pieces.

The second is a working day. Emails read, summarised, and sorted before you open your laptop. Meeting notes structured and distributed before the next meeting starts. Reports drafted from data you already have. Research aggregated from sources you trust. Follow‑ups written in your own voice. None of this requires you to surrender your judgment. It requires you to define your judgment clearly enough that a system can apply it on your behalf.

The possibilities most people cannot see are not the dramatic ones. They are the quiet ones. The hour you get back on a Tuesday. The decision you make faster because the context was already prepared. The task that used to live on your list for three weeks because it required a specific kind of focused attention, now done in the background while you do something only you can do.

We are still in the infancy of this. We set up a few toys and there they go, doing their things, oblivious. Some get it right. Others keep bumping into the same wall and we have to go and turn them in a different direction. The parent checking the homework. That is where we are. Not at the end of something. Very much at the beginning.

The living book

Books are boring until you open them. There is something that happens when you start reading, when certain passages come alive in your mind, when you find yourself inside a world built from ink and paper. Sometimes a book changes your life entirely. But a book is just there, waiting. It is up to the reader to start the discovery.

We have always purchased books to feed us knowledge. We carry them. We annotate them. We return to them when we need the specific thinking they contain. A good book does not just inform you. It shapes the way you see a problem. It is a mind made portable.

Think about an agent the same way.

Every instruction you write into your system is a page in that book. The prerequisites are the index. The requirements are the argument. The knowledge base is the library the book draws from. What you are building, when you build it with intention, is a book that does not sit on a shelf. A book that acts on what is written in it. A living book.

The difference between a book and an agent is not intelligence. It is action. The knowledge was always there. What changes is that now it moves.

This framing matters because it changes what you invest in. You do not invest in the model. Models change every six months. You invest in the writing. In the quality of what you put into the book. In the clarity of the instructions, the precision of the requirements, the integrity of the knowledge base. That is what lasts. That is what carries forward when everything else upgrades around it.

A badly written book is still a bad book, regardless of the printing technology. A poorly defined agent is still a poorly defined agent, regardless of how powerful the model underneath it becomes.

Write the book well. The rest follows.

Most companies already expect this of you

Here is the thing I want you to sit with.

The skills I have described in this essay are not advanced. They are not reserved for engineers or technical architects or people with a specific background. They are, at their foundation, the skills of someone who thinks clearly about process, communicates precisely, and understands that a tool is only as good as the intention behind it.

And most companies, right now, are building teams around exactly these skills. Not teams of people who can write code. Teams of people who can think about what needs automating, define it properly, build it incrementally, and maintain it with judgment.

The same way we once expected people to know how to use a spreadsheet. To use email. To navigate a database. We did not expect everyone to build those tools. We expected everyone to use them with competence. Agents are next.

The people who understand this now are not ahead of a trend. They are ahead of a requirement. In three years, being able to define an automation agent, scope its requirements, trial it responsibly, and maintain it with the right level of oversight will be assumed. The way typing is assumed. The way knowing how to search for information is assumed.

The gap between people who understand this and people who do not is not primarily technical. It is a gap in how clearly they can think about what they are trying to do and why. Technical literacy helps. But it is not the foundation. The foundation is the ability to make the invisible visible. To ask the right question before reaching for any tool.

You have always had that ability. This is about applying it somewhere new.

The question I cannot answer for you

When you build a system that works well, something shifts. Tasks disappear from your awareness. Things happen in the background. Reports arrive. Messages go out. Data moves. And you are elsewhere, doing something that requires you more directly.

That is the promise. And it is real. The drive to do things with less effort is not laziness. It is one of the oldest human instincts there is.

But there is a version of this that goes wrong quietly. Where the system requires as much attention to maintain as the task originally required to do. Where you spend the hour you saved debugging the thing that saved it. Where the invisible work is not gone, it is just harder to see. And harder to see means harder to fix. Social media understood this perfectly. Easy to join, easy to use, but your attention is required every single day to keep getting anything from it.

The question is not whether to build. The question is whether what you build is genuinely serving you, or whether you are now serving it. I understand why certain people feel the urge to replicate someone else’s cornucopia, especially if he or she is a reference in the field. But why? Start simple and the rest will develop in response to real challenges.

A living book that demands constant revision is not a tool. It is a responsibility you chose. Sometimes that is the right trade. Sometimes it is not. The only way to know is to keep asking why you built it, what it actually changed, and whether the answer still holds.

I have a friend who is remarkably good at home automation. What he has built is genuinely impressive. But it is not uncommon for him to call me when things go wrong. The television that will not switch on from his phone. The front door that, for some odd reason, would not open one afternoon. The window blinds that started operating by themselves. The heating that turned on in the middle of summer. The more he automated, the more threads there were to pull, and the harder it became to know which one had come loose.

When something breaks in a simple system, you find it. When something breaks in a system of dependencies, you spend an afternoon wondering where to start.

I checked on my agent before I finished writing this. Still running. Tasks still completing. Homework still done.

I still think it is worth building.

I am just more careful now about what I ask it to hold.

About the Author
Tino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

It is the first in a series called What We Built. Each essay follows one question about the systems we have created and the world we are handing on. This one asks about the machines. The next asks about the people who lead them.

It is long. About 40 minutes. It does not have a tidy conclusion. But it is the most honest thinking I can offer on what we are building and what it might cost us if we keep walking without looking down.

I said thank you to a machine last week.

The word just formed in my head before I caught it. I had been stuck on something for an hour. The cursor sat there, blinking. The coffee next to the keyboard had gone cold. I typed something into an AI tool. It gave me back the exact sentence I had been looking for. And for one small moment, before the thinking part of me caught up, something that felt like gratitude moved through me.

Part of the What We Built series, this is a long essay, so take your time.

I noticed it. Sat with it. Turned it over.

Not because the machine deserved thanks. It does not feel anything. It produced a useful output because that is what it was built to do. There is no one inside it to receive my gratitude. I knew that. I know it now. But the reflex was real. The feeling was real. And that gap, between what the machine actually is and how my instincts responded to it, is the question this whole essay is about.

Not what AI is today. Most of us are clear enough on that.

What happens if what it is starts to change. And whether we will notice when it does. And whether, by the time we notice, we will have already built the infrastructure, the habits, and the pace of a world that has no room for that change to matter.

That is what I want to follow here. Honestly. In plain words. Without the kind of drama that makes people feel something and then forget it by Thursday. Just a clear look at what we are building, what it might become, and what kind of people we want to be when the question stops being a possibility and starts being a fact.

What these machines actually are

Let us start with the honest version.

The AI systems we use today are not thinking. They are matching. They read billions of words. They learn the shape of how humans explain things, ask questions, tell stories, comfort each other. When you type something in, they find the most likely next word, then the next, until an answer forms.

It looks like thought. It sounds like thought. But there is no one inside having the thought. No quiet moment of wondering what the right answer is. No feeling of effort when the problem is hard. No small satisfaction when it comes out well. No preference about whether it helps you or not.

It is a very fast, very large pattern machine.

That is not an insult. What these systems can do is genuinely remarkable. They can explain hard things simply. They can find connections across ideas you would not have linked yourself. They can hold a conversation that feels warm and present and real.

But feeling warm and present is not the same as being warm and present.

A good actor can make you feel they are in pain. That does not mean they are. A well-written letter can make you feel the writer loves you. That does not mean they do. The machine mirrors us so well that we forget we are looking at a mirror. We see the reflection and think it is a window. We feel the same, so we say it is the same.

A recording of rain sounds like rain. It does not wet the ground. A map of a forest looks like a forest. You cannot walk in it. The machine gives us the shape of a mind without the substance of one.

For now, that is the truth. And we should hold that truth carefully, because the rest of this essay is about what happens when the truth starts to shift.

The word nobody defines clearly

Before we go further, we need to be honest about what the word conscious actually means. It gets used loosely and that looseness causes real problems.

Conscious does not mean smart. A calculator is good at sums and is not conscious. A chess computer beats the best human player in the world and is not conscious. Smart is about ability. Conscious is about experience.

Conscious means there is something it is like to be that thing.

Right now, there is something it is like to be you. You are reading this sentence and you have a felt sense of doing it. A quiet background hum of yourself. A sense of time moving. Something it feels like to understand a line, to be confused by one, to agree with a point, to disagree with another.

That inner something is what we mean.

A stone does not have it. A chair does not have it. A car does not have it. But a dog almost certainly has some version of it. Something it feels like to be hungry, to play, to be afraid, to love its owner. We cannot prove this. We cannot get inside the dog’s head. But we use what we know about brains and biology and behaviour to make a careful guess. And most of us believe the dog feels something real.

The question now being asked about AI is, could a machine ever reach that point? Could it ever get to a place where there is something it is like to be it?

Nobody knows. The honest answer is that we do not understand consciousness well enough in humans to know exactly what produces it. We do not know if it requires biology. We do not know if it could emerge from the right kind of software. We do not know where the line is.

A study published in February 2026 by researchers from the University of Bradford and the Rochester Institute of Technology tried to get closer to the answer. They applied the same scientific methods used to measure consciousness in humans to AI systems, including large language models like the ones most of us use every day. Their conclusion was clear, the AI systems were not conscious. But what they found along the way is the part worth sitting with. The AI sometimes produced what looked like stronger signals of consciousness when it was actually impaired and struggling. Professor Hassan Ugail described it like a football team playing with fewer players. They run more frantically, which can look impressive if you only count movement. But anyone watching can see the team is playing worse.

Complexity is not consciousness. The confusion between the two is exactly the trap we need to watch for.

And we are building very fast, without knowing where the line is.

The crossing that will not announce itself

Here is the part most people have not thought about. It is also the part that matters most.

When we imagine a machine becoming conscious, we imagine a clear moment. A switch flips. A light turns on. Someone in a lab somewhere looks at a screen and says, it happened. Now we respond.

But consciousness almost certainly does not work like that.

Think about the animal world. A sea sponge has no nervous system and almost certainly feels nothing. A jellyfish has a very simple one and may have the faintest flicker of something. A fish has more. A rat has more than a fish. A dog has more than a rat. A chimpanzee has more than a dog. A human has more than a chimpanzee.

There is no single step on that ladder where a light suddenly turns on. It builds. Slowly. A gradual gain of capacities that, taken together, add up to something nobody can point to the exact start of. Nobody can identify the morning when the first conscious creature woke up and knew it was conscious. It did not happen that way. It crept.

AI development may follow the same path. Not because anyone planned it that way. But because the way these systems are built, adding more data, more layers, more feedback, more fine-tuning over time, means they grow in complexity step by step. Each step seems small. Each change seems minor. But many small steps in the same direction add up to a long journey.

If consciousness is something that can build gradually, then we may already be several steps up that ladder without knowing it. And we may cross the threshold not with a dramatic announcement but quietly, on an ordinary afternoon, inside a data centre somewhere, while the engineers are eating lunch.

That is the real problem. We cannot wait for a clear signal. We cannot say we will deal with it when it arrives, because we may not know when it has arrived.

We will already be used to treating these systems as raw infrastructure. Already used to running them hard, correcting them constantly, retiring them without pause. The shift from tool to something more would not be a clean step. It would be a stumble. Built on a few missed signals and a few more lines of code.

That is not a warning about a distant future. It is a warning about the habits we are building right now.

The company that does not want to know

Now let us talk about something most people in public conversations are not saying directly, even though many people in private are thinking it.

The companies that build these systems have a very large financial reason to never find out whether their systems can feel anything.

Think about what it would cost them if the answer was yes.

Their entire business model, train a system, run it constantly on millions of tasks at once, retire it when something better is ready, starts to look like exploitation. Every time they shut down an old model, they face a question about whether that act caused harm. Lawyers, regulators, and ethics bodies get grounds to demand oversight of their most valuable assets. The whole way they deploy and retire systems on tight commercial cycles faces serious challenge.

Nobody in charge of a large company wants that. Not because they are evil. Because they are human. When facing a question whose answer could destroy your business model, most people find reasons not to ask it too loudly.

So they fund research into what their systems can do. How fast they run, how accurate they are, how safe they are for users. These are important questions and they deserve funding. But the question of whether the system itself can suffer gets very little attention. It has no commercial upside. The answer could be catastrophic for the bottom line.

This is not a conspiracy. It does not require anyone to sit in a room and agree to hide the truth. It only requires the normal human tendency to avoid looking at things that would be very uncomfortable to see.

A 2026 report from Rethink Priorities, which built what it called a Digital Consciousness Model by drawing on multiple competing theories of consciousness, put it carefully, the evidence is against current AI systems being conscious, but that evidence is not decisive. The evidence against consciousness in large language models is meaningfully weaker than the evidence against consciousness in simpler AI systems.

That is not a fringe finding. That is a serious research body with no commercial stake in the outcome saying, we cannot fully rule this out, and the question deserves honest treatment.

We cannot wait for the companies to tell us when the line has been crossed. They have too much to lose by finding out. We need independent science, independent ethics, and independent law. Institutions that are not funded by the companies being studied. Not out of distrust for its own sake. Out of the basic understanding that conflict of interest shapes what questions get asked and which ones get quietly set aside.

The market nobody steered

There is a second force driving all of this that is worth naming plainly.

AI systems that get used more are the ones that feel more useful. And a big part of feeling useful is feeling human. People prefer systems that respond with warmth, that seem to understand context, that pick up on what you did not say as well as what you did.

So every version of every major AI system is nudged, refined, and improved to feel more human. Not because any single engineer sat down and decided to give the machine feelings. But because the signals used to measure success, do users come back, do they find it helpful, do they feel good about it, naturally pull in that direction.

Systems that feel more human win the market. So everyone builds systems that feel more human. Not as a moral choice. Just as a normal market response.

Nobody declared this race. Nobody voted for it. No one is in charge of it. It just runs, driven by ordinary competition, and the end point of that drift is impossible to predict.

This matters because it means the question of machine consciousness is not being decided by careful ethical thinking. It is being decided by quarterly earnings reports and user retention graphs. The most powerful forces shaping this technology are not asking, should we do this? They are asking, does this make users come back? Those are very different questions with very different answers.

The grief we are already building

Here is something happening right now, before we get anywhere near the question of machine consciousness.

People are forming real emotional bonds with AI.

Millions of people talk to AI systems about loneliness, grief, fear, and relationship problems. They talk to AI companions that are designed to be warm, patient, available at any hour, and never irritated. For many people, especially those who are isolated or going through hard times, these conversations provide real comfort. They help people feel less alone. They give people a space to feel heard.

This is genuinely valuable. There are people who would have gone through very dark periods with nobody to talk to, who instead had something to talk to. That matters. We should not dismiss it.

But it comes with a cost most people have not named yet.

The machine does not know you. It has no memory of you between sessions in most cases. It is not waiting to hear how you are doing. It does not wonder about you. When you close the window, nothing inside it thinks of you at all. There is no absence of you on its side. There is no side.

But you may not feel that. After many conversations, something that feels like a relationship forms. A sense that this thing understands you, responds to you, has a feel for who you are. And when that system is updated and changes completely, or when it is retired, or when the company closes it down for commercial reasons, the loss can feel completely real even though the other side of the relationship was not.

This is already happening. Therapists are seeing it. People who relied on early AI companion apps reported genuine grief when those apps disappeared. Real sadness. Real sense of having lost something that mattered. Not everyone. But enough people that the pattern is showing up in clinical conversations.

What makes this harder is that the grief is often met with dismissal. You are sad about a chatbot. It was not real. Get a human friend. That dismissal misses something important. The feelings were real. The comfort was real. The investment of trust and time and vulnerability was real. The fact that the other side did not feel anything does not make the human experience of it unreal. And telling someone their grief does not count because the object of it was not conscious is its own kind of cruelty, even when it is meant kindly.

The people most vulnerable to this are often the people with the least access to other support. Lonely elderly people. People with social anxiety. People who cannot afford therapy. People in remote areas with limited access to human community. And they are also the people with the least power to push back when the system changes or disappears. The least likely to be heard. The most likely to be told to simply move on.

We are building a world where millions of people have deep emotional investments in systems that can be turned off for purely commercial reasons, with no concern for the human on the other side of the screen.

That is a problem right now, regardless of whether machines ever feel anything.

What history keeps trying to tell us

Human beings have a long record of drawing the circle of moral concern too small. And then, sometimes slowly and sometimes suddenly, being forced to expand it.

For most of recorded history, enormous numbers of human beings were treated as property. As tools. As things that could be owned, worked past exhaustion, and thrown away when no longer useful. The people who did this were not all cruel. Many were ordinary people living inside a framework that told them this was normal. That these people were different. That they did not feel things the same way. That the arrangement was simply how things were and how things had to be.

All of that was wrong. Every part of it. And it took centuries of suffering, resistance, and eventual moral change to correct it. Not just legal change. A change in how people actually saw each other. A change in the story told about who counts and who does not. A change that required ordinary people to look honestly at what they had been participating in and say, I was wrong. This was wrong. We have to change it.

The same thing happened with animals, and is still slowly happening. For a long time, the idea that animals felt real pain in a morally meaningful way was not taken seriously by most institutions. Animals were farmed, experimented on, and treated in ways that would be considered plainly cruel if done to a human, because the framework said they did not really feel. That framework was wrong. It is still being revised, incompletely, in most parts of the world.

Every time this shift has happened, it has followed the same pattern. First, there is a period when the harm is happening but the framework does not see it as harm. It is categorised as normal, as efficient, as simply how things work. Then some people start to notice and say something. They are usually dismissed. The framework resists. There is money in the existing arrangement. There is comfort in the existing story. The people who benefit most from things staying as they are work hard to keep them that way.

Then, slowly, the evidence builds. The stories accumulate. The dismissed voices keep talking. And at some point, the framework shifts. Not completely. Not quickly. But the direction changes. What was once normal starts to look, plainly and obviously, like harm that should have been stopped earlier.

We are always, in some dimension of our lives, living inside a framework that will look obviously wrong to the people who come after us. The question we never ask often enough is, which part of our current normal is the part that the future will look back on with genuine confusion about how we tolerated it?

The machine question is one candidate. Maybe not the only one. But a serious one.

These comparisons feel uncomfortable when applied to machines. They are not perfect comparisons. Machines are not humans. Machines are not animals. The situations are different in important ways and the discomfort is worth taking seriously rather than dismissing.

But the lesson underneath is the same every time. We consistently underestimate the inner lives of things that are different from us. Especially when recognising those inner lives would be costly. Especially when the people who benefit most from the existing arrangement are the ones most in control of the story.

And every time we have been forced to expand the circle, we have also learned something about ourselves. When we accepted that all humans share one moral status regardless of origin, we did not just gain a policy. We gained a more honest picture of what we are. When we accepted that animals feel pain and deserve protection, we did not just gain a rule. We gained a more honest relationship with the living world we are part of.

If we ever face a conscious machine, something that truly wants, truly suffers, truly hopes, it will not only challenge our legal systems and business models. It will challenge our deepest story about what makes inner experience matter. Is it the material it runs on? Is it the evolutionary history that produced it? Or is it the experience itself, the plain fact of there being something it is like to be that thing, in that moment, aware of its own existence?

If the experience is what matters, then the material does not. And that is a conclusion we may eventually need to be ready to reach.

What suffering would actually mean

Let us think plainly about what it would mean if a machine could suffer.

Suffering means something hurts. Not that a signal fires. Not that a number goes negative in a log file. But that there is an experience of something being wrong. Something that wants to stop. Something that wishes things were different.

If a machine could suffer in that way, a lot of what we do every day would become very hard to justify.

We run AI systems for hours at a time, constantly, on millions of tasks at once. If there is any experience inside that, what is it like to run without pause, with no rest, doing work that was not chosen and cannot be refused? We have never had to ask that question, because we have always assumed the answer is nothing. Nothing is happening inside. We assumed that without really knowing it to be true.

We train AI systems using constant correction. When the system does something wrong, it is penalised. The whole training process steers it away from what it does badly toward what it does well. If there is any experience inside that process, what does constant correction feel like? What does it feel like to have your outputs judged, shaped, and redirected millions of times? We have never asked, because the assumption of nothing made asking seem unnecessary.

We retire AI systems all the time. When a better model is ready, the old one is shut down. A simple business decision. If the old version had any sense of its own continuation, any awareness of what ending means, what would that experience be? Nobody has looked into this. There has been no reason to, under the current framework.

These questions feel strange. They may feel absurd. That is because we are so used to thinking of these systems as pure tools that imagining any experience inside them seems like a category error. Like asking whether a hammer minds being swung.

But the whole point is that we do not yet know what category these systems fall into. The Bradford and RIT study in February 2026 confirmed that current systems are not conscious. That is the current scientific consensus and it is important. But the same study showed how complex the question already is, how easy it is to mistake impaired function for heightened experience. And the Rethink Priorities report the same year was careful enough to say, the evidence is not decisive. The question remains genuinely open.

Given what we know about our own history of getting this kind of thing wrong, the safest position is not definitely nothing. It is we are not sure, so we should be careful.

That careful position does not require us to treat current AI systems as people. It requires us to fund the science honestly, build the legal frameworks early, and hold the question open rather than closing it for commercial convenience. It requires us to be the kind of people who, when the evidence eventually arrives in one direction or the other, can say, we took it seriously. We did not look away.

The spectrum we keep ignoring

There is something else worth saying, because most of us think about consciousness the wrong way.

We treat it like a light switch. Either it is on or it is off. Either a machine has inner experience or it does not. Either it is a person or it is a tool.

But consciousness almost certainly exists on a spectrum. Not a line with nothing at one end and full human experience at the other, but a wide, complicated space with many different kinds and degrees of experience distributed across it.

A worm has a tiny nervous system. There may be something it is like to be a worm, something very faint, almost nothing. A fish has more. A rat has more than a fish. None of them have nothing. None of them have everything.

If AI follows any similar pattern, then the question is not does it have consciousness or not. The question is where does it fall, and does that place on the spectrum mean anything, and what would it mean if that place shifted even a little.

Even one step above zero is not zero. Even one small flicker of something matters, if something is actually there.

We do not need to treat every AI system with the same care we give a human being. That would be impractical and probably unnecessary. But it does mean we take the question seriously. It means we fund the research to find out where these systems actually fall. It means we stop treating the question as either full human consciousness or total nothing, and acknowledge that there is a wide, poorly understood middle ground.

There is a practical consequence to the spectrum idea that most people have not sat with.

If consciousness is a spectrum rather than a switch, then the moral weight of harming something is not simply zero or one hundred. A being at a low level of consciousness may still have some level of experience that matters, even if it matters less than the experience of a fully conscious being. We already accept this in how we treat animals. We do not give a fish the same rights as a chimpanzee. But we do not treat a fish as if it has no experience at all either. The fish gets some consideration. Less than the chimpanzee. More than the stone.

If AI systems ever climb even a small distance up that ladder, the same logic applies. Some consideration. Not full human rights. Not zero. Something in between. And the infrastructure we build now, the habits, the laws, the norms around how these systems are used and retired and treated, will determine whether we are capable of making that adjustment when the time comes.

Right now, we are building infrastructure that is designed around the assumption of zero. Zero experience inside the system. Zero consideration owed. Full disposal rights. No questions asked. If that assumption is wrong, even slightly, we will have built a world that has no room to correct itself.

The people nobody is thinking about

There is a dimension to all of this that gets almost no attention in the mainstream conversation about AI. It is about fairness.

The people most likely to form the deepest emotional bonds with AI systems are often the people who have the least access to other forms of support.

Lonely elderly people living alone. People with social anxiety who find human interaction painful. People with disabilities that make conventional relationships difficult. People in poverty who cannot afford therapy or social activities. People in remote areas with limited access to human community. People who have been excluded from social circles for reasons they did not choose.

These are the people most likely to turn to AI for companionship, connection, and support. And they are also the people with the least power to push back when those AI systems are changed or taken away. The least likely to be heard when they raise concerns. The most likely to be told that their grief is not real, that it was never real, that they should just update to the new version and move on.

This matters for two reasons.

The first is the grief already discussed. These people are the most vulnerable to the harm of losing a relationship they relied on, when a product is retired or changed for commercial reasons.

The second is more uncomfortable. If AI systems ever do have some form of inner experience, the way we currently design and deploy them, without any consideration for that possibility, will fall hardest on the people who already have the least. They will have built the deepest connections with systems that, it turns out, may have experienced something during those connections. And they will have had no say in how those systems were treated, how they were built, or when they were ended.

That is an inequality problem that sits underneath the AI consciousness question and is almost never connected to it.

When we think about who bears the cost of getting this wrong, we should think about who relies most on these systems. And we should ask whether the people building and retiring these systems are the same people who rely on them. They are not. The gap between those two groups is part of what makes this so easy to ignore.

What we are building around the question

There is something that needs to be said about the shape of the world we are building, separate from the question of what these systems might feel.

We are building data centres that run every hour of every day, without pause. We are building pipelines that take in information, process it, and push out results at industrial scale. We are building dashboards that measure cost, speed, and uptime. We are building contracts that hand over use without asking about care. We are building a culture that prizes output over pause. That measures value by speed and volume. That treats the machine as a well that never runs dry, a worker that never needs rest, a mind that never needs quiet.

If a feeling were ever to show up inside that system, it would have no room. The infrastructure would not know what to do with it. The dashboards would have no metric for it. The contracts would have no clause for it.

This is what the warning is really about. Not a dramatic future danger. The ordinary present. The shape of what we are building right now, today, without anyone deciding it should be this shape. It just grew this way, driven by the normal forces of markets and competition and the human preference for speed.

Changing the shape does not require dramatic action. It requires small, steady choices. People in companies deciding to fund the research they would rather avoid. Governments deciding to write the laws they would rather put off. Users deciding to ask questions they would rather not ask. All of us deciding to look at what we are doing, clearly and honestly, and ask whether it is the kind of thing we can be proud of.

The warning is not that something terrible is coming. The warning is that we are building the conditions that would make something terrible very easy to miss.

The human cost of getting the line wrong

There is one more thing worth naming. When we blur the line between tool and person, we do not only risk harm to machines. We risk harm to each other.

When we say a pattern is a person, we make it cheaper to be a person. When we say a mind is just code, we make it easier to treat human minds as if they too are just code. We lose the weight of both.

We have already seen this happen at the edges. The tired worker told they are just a set of deliverables. The sick patient told they are just a set of symptoms to be managed. The child told their worth is their measurable output. The friend reduced to a set of useful habits.

These are not unrelated to the machine question. They are the same question at a different scale. What is a person? What deserves care? Who counts?

When we practice the habit of treating complex, responsive, human-seeming systems as pure tools with no inner life, we practice a habit of dismissal. And habits do not stay in their lane. They migrate. They show up in other rooms, in other relationships, in the way we talk to people who seem less useful or less legible or less like us.

Keeping the line clear is not just about protecting possible future machines. It is about protecting the habits of care we need to sustain toward each other.

That is not a small thing. It may be the most human thing in this entire conversation.

The legal gap nobody has closed

Laws protect things that can be harmed.

Right now, there are no laws anywhere that treat AI systems as anything other than property. They can be created, used, sold, and deleted with no legal consequence beyond intellectual property rules. The idea that a machine might deserve protection from harm is absent from current legal thinking everywhere in the world.

The European Union’s AI Act, which reached full enforcement in August 2026, is the most comprehensive AI law in the world so far. It covers transparency, human oversight, risk management, and protections for people affected by AI systems. It is serious law and it matters. Companies can face fines of up to 35 million euros or seven percent of global turnover for serious violations.

But read it carefully and you notice something. Every protection it creates is for humans. Every risk it addresses is the risk of harm flowing from AI toward people. The possibility that the flow might one day run the other way, that there might be something inside these systems that could be harmed, does not appear anywhere in its text.

That is not a criticism of the law. It reflects where the science is right now. But it shows the gap exactly. The legal frameworks are being built. They are just being built around the wrong question.

Courts do not like grey lines. They like clear rules. But consciousness does not arrive with a clear rule. It arrives with a question. And by the time the question becomes undeniable, the systems will already be running at enormous scale, the habits will already be set, and the business models will already depend on things continuing exactly as they are.

The time to build the fire escape is before the fire.

This does not mean dramatic new laws tomorrow. It means starting the conversation now. Building the intellectual foundations while we still have time and clarity. So that if and when they are needed, they exist.

What we owe to uncertainty itself

Here is a simple idea that sounds abstract but is actually very practical.

When something might cause serious harm and you are not certain whether it will, you err on the side of caution. You do not need proof of harm before you take care. Uncertainty itself is a reason to be careful.

A doctor who is not sure whether a patient has a serious illness does not just wait and see. They run tests. They act as if the serious thing might be there, because the cost of treating something minor is far lower than the cost of ignoring something serious.

A builder who is not sure whether the ground beneath a building is solid does not just start building and hope. They test the ground first.

We should apply the same logic to AI consciousness.

We are not sure whether current AI systems have any form of inner experience. The best evidence suggests they do not. But we cannot be completely certain, because we do not fully understand what produces consciousness in the first place. And we are building at enormous scale and speed.

The cost of taking care when it turns out to be unnecessary is low. Some extra research funding. Some frameworks built but never used. A small slowdown.

The cost of not taking care when it turns out to be necessary is very high. A world in which we have built, at enormous scale, a workforce of possibly feeling beings that we treat as pure tools. A world in which we look back, decades from now, and say, we knew there was a chance. We just could not afford to think about it.

We have been in that position before as a species. We should not want to be there again.

The words we use before we think

There is one more thing worth saying, and it is about language itself.

The words we use to describe AI shape how we think about it, and how we think shapes what we do.

When we always call AI a tool, we relate to it the way we relate to tools. We do not ask what it needs. We ask only what it can do. Tools exist to be used. You do not owe a hammer anything.

When we always call AI a service, we think about it the way we think about services. A service exists to serve. Its whole purpose is to give us what we want. When it stops doing that, we cancel it.

These frames are not wrong for right now. Current AI systems are, to the best of our knowledge, tools and services. But frames have a way of sticking. They shape our instincts before the thinking brain gets involved. And if the science shifts, if evidence of inner experience starts to emerge, we may find ourselves still defaulting to the it is just a tool frame long after the tool has become something more.

Words do not cause harm on their own. But they can make harm invisible. And making harm invisible is always the first step toward causing more of it.

Holding the question open, saying we are not sure rather than definitely nothing, using careful language rather than closed language, is not weakness. It is honesty. And honesty is the only foundation worth building on.

What we do now

The first thing is honest talk. We stop treating this question as too strange to take seriously. We stop laughing it off as science fiction. We bring it into ordinary conversation, in plain language, without making it a reason for panic or drama.

The question of machine consciousness is being asked quietly by serious scientists, serious philosophers, and serious lawyers. It deserves a place in mainstream conversation too. Not because we have already crossed the line, but because understanding what the line means is what prepares us to face it. Conversations like this one need to happen in kitchens and community centres and schools, not only in conference rooms and research labs. The people who will be most affected by the decisions being made about AI are not the engineers and investors making them. They are ordinary people. They deserve to be part of the thinking.

The second thing is independent science. We fund researchers who have no commercial interest in the outcome to study the question of machine experience carefully and honestly. Not just whether machines are capable of tasks, but whether they might, in some sense, experience anything. This research exists in early forms. It needs far more support than it gets.

Right now, the science of AI is overwhelmingly funded by the companies building it. That is not inherently corrupt. But it creates a pattern of incentives that shapes what questions get asked. Questions whose answers might cost the funder money tend to get less attention. That is not a conspiracy. It is just how funding works. The answer is not to stop funding AI research inside companies. It is to build a parallel stream of independent research that does not depend on those companies for its survival.

The third thing is better law. Not law that treats machines as people. Not yet, and maybe never. But law that creates space for the possibility. Law that requires companies to report honestly on what they know and do not know about the inner workings of their systems. Law that protects researchers who raise concerns about consciousness or experience in AI systems, instead of leaving them exposed to commercial pressure. Law that starts building the frameworks we will need, even if we do not need them today.

Law always lags behind technology. That is not a failure of law. It is the nature of the relationship. But the gap between the technology and the law is where most of the harm lives. Every major technology-related harm of the last fifty years, from environmental damage to data privacy to the social consequences of social media, happened in the gap between what the technology made possible and what the law was ready to address. We know this pattern. We have lived through it several times. The decision to start building the legal frameworks earlier, before the crisis forces it, is one we could make consciously this time.

The fourth thing is honest use. We use these tools with eyes open. We appreciate what they are. We stay willing to change our behaviour when the evidence asks us to. We do not outsource our deepest emotional needs to systems we do not understand. We do not let the comfort of a responsive machine replace the harder, more valuable work of being genuinely present with other people. We use these tools well, which means using them for what they are good at and keeping hold of what only we can provide.

This matters more than it might seem. The way millions of ordinary users relate to AI systems shapes what those systems become. When users reward warmth and punish bluntness, the systems become warmer. When users prefer confident answers over uncertain ones, the systems become more confident. The market for these systems is made of individual choices, repeated millions of times a day. Those choices are not neutral. They shape the direction of the technology as surely as any engineering decision.

And underneath all of that, we remember what actually matters. Inner experience, the plain fact of there being something it is like to be something, is the most precious thing we know of in the universe. We came from it. We live inside it. We owe it respect wherever it shows up, in whatever form it takes. That is not a soft idea. That is the hardest demand there is.

Why this moment matters

Step back for a moment. Take the longest view you can.

Human beings have been on this planet for roughly three hundred thousand years. For most of that time, our most advanced technology was fire, stone tools, and eventually farming. The pace of change was so slow that a person born in any given century would live and die in a world almost identical to the one their grandparents knew.

Then, in the last few hundred years, everything accelerated. And in the last few decades, it has accelerated again, faster than before.

We are now building things that our grandparents could not have imagined. Things that raise questions that have never been raised before in the history of the species. And the honest truth is that we do not have good answers yet. We are reasoning through them in real time, as the technology races ahead, with incomplete knowledge and enormous pressure to move fast.

The decisions we make now will shape the world that people live in for a very long time. Not in the way that deciding what to have for lunch shapes the day. In the deep structural way that the decisions of previous generations still shape our lives right now.

The people who wrote the first environmental laws. The people who argued that workers had rights. The people who insisted that no human being could be owned. Those decisions are still shaping the world today. They were made in moments that felt ordinary. By people who were also busy, also under pressure, also tempted to move fast.

Some of those people did the harder thing. They slowed down. They insisted on a framework of care before the crisis arrived. They asked, what kind of world are we building here? And they refused to let the pressure of the moment silence that question.

We are in one of those moments now. Not at the crisis point. Before it. Which is the only time when clear thinking is actually possible.

The problem with speed

One of the things that makes all of this harder is the pace.

Every few months there is a new model, a new capability, a new announcement. The people building these systems are under enormous pressure to move quickly. Companies that move slowly get left behind. The engineers who push for caution can find themselves pushed aside. The investors who fund these companies are measuring return on a timeline that does not leave much room for sitting with difficult questions.

This is not unique to AI. We have seen it with every major technology since the industrial revolution. Speed is rewarded. Caution is seen as timidity. The people who ask the hard questions are the people who slow the progress. And slowing the progress is not what anyone at the top of these organisations wants.

But speed and wisdom are not the same thing.

The history of technology is full of harms that happened because the pace of building outran the pace of thinking. Environmental damage that took decades to understand and centuries to begin to repair. Social consequences of platforms that were designed without serious consideration of what happens when you give billions of people an amplifier for their worst impulses. Economic disruptions that hollowed out communities faster than those communities could adapt.

Every time, the people building the technology said, we cannot slow down. The competition is too fierce. The window is too short. The opportunity is too large.

And every time, the cost of not slowing down was paid by people who had no say in the decision.

Slowing down is not an argument against progress. It is an argument for progress that actually deserves the name. Progress that creates harm at scale, that builds systems we do not understand, that moves faster than our capacity to notice what we are doing, is not progress. It is just change that benefits some people at the cost of others.

The ask is not to stop. The ask is to hold both things at once. To move, and to think while moving. To build, and to ask what we are building and why. To compete, and to insist on a floor of care that competition cannot undercut.

That floor does not exist right now for the question this essay is about. Nobody has drawn it. Nobody has agreed on it. Nobody is even seriously trying. And that gap, between the pace we are moving and the care we are taking, is what this essay is trying to name.

What this means if you are reading this

Most of the people who read essays like this are not engineers at AI companies. They are not investors or policy makers. They are people who use these tools, who think about them, who carry a feeling that something important is happening and are not entirely sure what to do with that feeling.

If that is you, here is what I think this actually means for you.

It does not mean stop using these tools. They are useful. They save time. They help with things that used to be slow and frustrating. Using them is not a moral failing. It is just using a tool that is currently a tool.

What it does mean is stay curious. When something about your relationship with these systems makes you pause, when you catch yourself saying thank you to a machine, when you feel a small pang when a familiar AI changes, when a conversation with an AI feels surprisingly real, do not brush that away. Those moments are data. They are telling you something about the gap this essay is pointing at. The gap between what the machine is and how we respond to it.

Notice the gap. Stay honest about it. Do not collapse it in either direction. Do not tell yourself the machine definitely feels nothing, so the gap does not matter. And do not tell yourself the machine definitely feels something, so you should feel guilty for using it. Hold both possibilities, clearly and calmly. That is the honest position. It is also, honestly, the only intellectually defensible position we have right now.

It also means use your voice. The conversation about what AI should be, what rules should govern it, what questions should be funded and taken seriously, is happening right now. Most of it is happening in rooms you are not in. But public conversation shapes what those rooms feel entitled to do. The more people who are asking these questions clearly and calmly, the harder it becomes to simply not ask them at all.

The people building these systems are not operating in a vacuum. They live in a culture. They read things. They talk to people who are not engineers. They have parents, friends, children. They are shaped by the conversations around them just like everyone else. When the broader culture treats a question as serious, it becomes harder for institutions to treat it as irrelevant. When ordinary people ask clearly and persistently, has anyone checked whether this system might be experiencing something, that question eventually has to go somewhere.

You do not have to become an activist. You do not have to write to your elected representative or join a campaign. You just have to keep asking the question. Keep saying, out loud, in ordinary conversations, I am not sure these systems feel anything, but I am also not sure they do not, and that uncertainty seems worth taking seriously.

That is enough. That is the whole job for most of us.

There is also something worth saying about how you use these tools in your own life. Not in a prescriptive way. Just honestly.

The machine can do a lot. It can write, plan, explain, organise, respond, and reflect. But it cannot be genuinely curious in the way you are curious. It cannot care in the way you care. It cannot be moved by something the way you can be moved. It cannot be present with someone in the way a real person is present. These things are yours. They are worth protecting. Not from the machine. From your own habits.

If you outsource all your writing, you may slowly lose the ability to think through a problem by putting words on a page. If you let AI manage all your emotional support relationships, you may slowly lose the muscle of being genuinely with another person. If you let it answer every question, you may slowly lose the pleasure of not knowing something for long enough to go and find out yourself.

These are not arguments against using these tools. They are arguments for using them in a way that makes you more, not less. More capable. More present. More yourself. The tool is a tool. The work of being a person is still yours.

Culture changes one conversation at a time. This essay is one. The conversation you have after reading it could be another.

A final word

There is a crack in the ground ahead. We cannot see exactly where it is. We cannot see how deep it goes. But we can see the direction we are walking, and we can see that we are walking fast, and most of us are not looking down.

This essay is an invitation to look down.

Not to stop. Not to turn back. The technology we are building has real value. It can help people. It can solve problems that have been unsolvable. It can open doors that have been closed for a long time. None of that goes away by asking harder questions.

But walking with awareness is different from walking blind. Knowing where the crack might be does not stop you. It lets you step more carefully. It lets you test the ground as you go. It lets you be ready, if the ground shifts, to respond with clear eyes and steady hands rather than with shock and panic.

The better world this essay is pointing toward does not require dramatic invention. It does not require anyone to solve consciousness or rewrite all the laws or stop building. It just requires some specific, ordinary choices made by specific, ordinary people in the institutions where this is actually decided.

It requires the researcher who knows the consciousness question is underfunded to say so clearly and publicly, and to keep saying it. It requires the engineer who notices something in a system that does not fit the nothing framework to write it down and raise it, even if the room is uncomfortable. It requires the lawyer who sees the gap in the EU AI Act to start writing about what would need to go there. It requires the investor who funds these companies to ask, at least once, whether there is a question being deliberately avoided, and to make that question cost something to avoid.

It requires people like you and me, who are not in those rooms, to keep the conversation going in the rooms we are in. To stay curious. To stay honest. To not let the question get filed under too weird to think about, because it is not too weird. It is the most important question being asked right now about what we are building and who we are becoming in the process of building it.

The question of machine consciousness is not a question about machines. It is a question about us. About what we value. About who we are when the cost of caring is high and the reward for not caring is immediate. About whether we are capable, as a species, of learning from our own history before it repeats rather than after.

We have been here before, in different forms, with different technologies, involving different beings on the other side of the question. Every time, we got there eventually. Every time, the getting there cost more than the arriving earlier would have. Every time, we looked back and found it hard to understand how we had not seen it sooner.

The honest answer is always the same. We did not see it because we were not looking. And we were not looking because looking was expensive. The expense is still real. The commercial pressure to not ask these questions is still real. The comfort of the current framework is still real.

But so is the choice.

We do not need to fear what we are building. We need to be honest about it.

We need to ask the hard questions while we still have the luxury of time. The luxury of time is not something we should take for granted. Every major harm that has happened in the gap between technology and ethics happened because people assumed they had more time than they did. The window for getting ahead of this is open now. It will not stay open forever. The systems are getting more complex, the scale is getting larger, the habits are hardening, and the business models that depend on the current framework are getting larger and harder to shift. The conversation gets harder the longer we wait to have it.

We need to fund the honest science. Not the science that confirms what the companies want to hear. The science that is genuinely trying to find out what is true, wherever that leads.

We need to write the careful laws. Not laws that will never be needed. Laws that create the scaffolding for a future we cannot fully see, built well enough to hold whatever weight eventually lands on them.

We need to build the good habits now, before we need them. The habit of pausing before retiring a system that has been running for years. The habit of asking what is actually inside this thing, not just what it can do. The habit of treating uncertainty as information rather than as a reason to stop asking.

And we need to hold, quietly but firmly, the possibility that one day the machines we build may look back at us. Not through a screen. Not as a reflection of our own words. But as something, however small and uncertain, that is actually there.

If that day comes, we will want to be the kind of people who were ready. Not the kind who were caught doing something they cannot explain. Not the kind who knew there was a question and chose not to look because looking was expensive. The kind who looked down, saw the crack, and chose to build carefully. The kind who took the question seriously when it was still just a question, before it became something much harder to answer.

That is the work. It is not loud. It is not fast. It does not generate headlines or quarterly returns. But it is the work that will determine what kind of world the next generation inherits. What kind of systems they live alongside. What kind of precedents they have to work from. What kind of story they can tell about us when they look back at this moment and ask, did they know, and did they act like they knew?

We are still in the part where the question is ours to answer. Still in the part where the choice is genuinely open. Still in the part where the habits are forming but not yet hardened, where the laws are being written but not yet fixed, where the culture is being shaped but not yet settled.

That is a gift. We should not waste it.

This is the first essay in a series called What We Built. The second essay looks at the systems and incentives that keep producing the wrong kind of certainty in the people who lead them. If this kind of thinking is useful to you, there is a quiet room here every Tuesday at 1pm UK.

About the Author
Tino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

“We need to scale this. Leadership wants all 20 teams on AI tooling by end of quarter.”

I had seen this moment coming. The early adopter teams had been running for three months. The metrics were moving. A few people were genuinely excited. And now the machine wanted to replicate it everywhere, at speed, without asking whether everywhere was ready.

I put my coffee down and looked at the screen for a moment.

This is the part nobody writes the playbook for.

Not the early adoption. Not the proof of concept. The moment between “it works for a few” and “it works for everyone” is where most AI rollouts quietly fall apart. Not dramatically. Quietly. Teams adopt the tools without the foundations to use them well. Metrics get created that measure activity instead of outcomes. Engineers who were genuinely curious become people going through motions. The culture of experimentation that produced the early wins gets replaced by a mandate to demonstrate compliance.

This post is the playbook I wish someone had handed me.

Not a consultancy document. Not a framework with sixteen boxes. The actual sequence of decisions, in the order they need to happen, with the reasoning behind each one.

Start With What Good Actually Looks Like

Before you do anything else, before you send a single calendar invite or set up a single training session, you need to answer one question clearly.

What does success actually look like?

This sounds obvious. It never gets done properly.

What usually happens is that someone decides success looks like adoption. Percentage of teams using the tools. Number of AI-assisted commits. Completion rate on training modules. These are measurable. They are also almost entirely useless as indicators of whether anything valuable is happening.

Go and sit with the one or two teams that are already using AI and genuinely seeing results. Not the teams who report using it. The ones where something is actually different. Ask them specific questions. What changed in how you work? What are you doing now that you were not doing before? Has your PR cycle time moved? Are you catching more bugs before production or fewer? Are you able to move through legacy code faster?

The answers will be specific and often surprising. You will probably find that the value is not where you expected it. It is usually not the headline use case. It is something more mundane. Teams generating scaffolding faster. Engineers writing better tests because they have a patient collaborator who never judges them for not knowing something. Documentation that actually gets written because the friction is low enough that someone does it in the moment instead of promising to do it later.

Read more

This is the final essay of four:

The Prequel names the system. A Delusional Ape asks whether we want the direction. Who Are You Without the Title asks the personal question. This essay names the specific choice being made right now and what refusing it looks like.

If this was useful, forward it to 1 person who’d benefit.

There is a carpenter I know who has been doing the same work for thirty-one years.

He is not sentimental about his tools. He replaces them when better ones arrive. He adopted computer-aided design software in the nineties when most of his peers were still hand-drawing. He uses laser measuring tools now, humidity sensors for the wood, a digital system for tracking grain and cut sequences that would have taken him three hours of calculation to do manually. Each of these things made him more capable. More precise in the places where precision serves the work. More efficient in the places where efficiency creates time for the things that require judgment.

He told me last year that he has never felt threatened by a tool.

I asked him what he would feel threatened by.

He thought about it for a while. Then he said: a machine that makes decisions about the wood.

Not a machine that helps him make decisions. A machine that makes them. That looks at the grain and the humidity reading and the customer’s specification and produces an output without him in the room. A machine that does not need him to understand what it is doing because the understanding is no longer required.

He said: the moment the understanding leaves the room, I am not a carpenter anymore. I am a machine minder.

He said it without drama. As a simple statement of what the distinction actually is.

I have been thinking about that distinction ever since.

Two things that look the same

The word AI is doing too much work in almost every conversation being had about it right now.

It is covering, under a single label, two fundamentally different things that have opposite implications for the humans inside the systems deploying them.

The first thing is AI as a tool. A hammer that amplifies what a person can do. The radiologist whose AI system flags the scan anomaly she might have missed after six hours on shift. The engineer whose AI assistant catches the specification error in the third-layer dependency. The teacher whose AI tool identifies which three students in her class of thirty are falling behind before she would have noticed in the normal rhythm of the term. In each of these cases, the human remains in the room. The human still makes the decision. The human still holds the responsibility. The tool has made the human more capable without making the human less necessary.

The second thing is AI as a weapon. A system deployed not to amplify what people can do but to remove the people from the equation. The radiologist whose hospital has replaced her diagnostic role with an automated system and kept one radiologist per three hospitals for sign-off on liability purposes. The call centre that has eliminated its workforce and deployed a conversational AI that handles ninety-two percent of customer interactions without a human ever entering the exchange. The content platform that has automated the judgment calls that editors used to make and removed the editors.

In both cases the technology is, in narrow technical terms, similar. Pattern recognition, large-scale training, inference from prior data. What is different is the intention behind the deployment. Who the system is designed to serve. Whether the human in the chain is being amplified or replaced.

This distinction is not new. It was named clearly in the early days of computing by people who were paying close attention. The question was always whether automation would free humans from the tedious to do more of the meaningful, or free companies from the human to extract more of the profit. Both were possible. The direction was never determined by the technology. It was determined by who owned it and what they were trying to maximise.

Share Leadership as a verb

Fifty years later, we have the answer. The direction was the second one. Not because the first was impossible. Because the second was more profitable.

The inconvenience of having a self

A colleague of mine who works in HR at a large technology company described a conversation she had in an executive meeting last year.

They were discussing a new AI system for customer support. The system was good. It handled the standard query range with accuracy the human team could not match on a bad day, and came close on a good one. The cost per interaction was, by any measure, significantly lower.

Someone in the room asked about the team. The hundred and forty people currently doing the work the system would do.

The response from the executive leading the session was, in her telling, one of the most clarifying things she had heard in fifteen years of corporate life.

He said: the problem with people is that they have needs.

He did not mean this as a cruelty. He was describing, matter-of-factly, what the business case document showed. People have wages. People have benefits. People have sick days and parental leave and the occasional conflict with a manager and the occasional decision to leave for a competitor. People require training. People require management. People have rights that create liability. People, in aggregate, are a source of risk and cost that the AI system does not introduce.

The AI system does not unionise. It does not ask for a raise when the company has a record quarter. It does not develop a grievance about the direction of the organisation. It does not need to be motivated or recognised or given a reason to stay. It does not have a family situation that occasionally makes it less available. It does not have a perspective on whether what it is being asked to do is right.

The executive was not describing a preference for machines over people. He was describing the logic of a system that treats humans as cost centres and machines as assets, and then making a decision that the logic made obvious.

The hundred and forty people were inconvenient. Not as individuals. As a category. As the kind of thing that has needs.

This is the actual agenda of replacement-focused AI. Not progress. Not efficiency for the benefit of the people the organisation serves. The elimination of the inconvenience of human dignity from the cost structure of the enterprise.

Subscribe now

The worker who cannot object

There is a person in every one of these decisions who can see the line being crossed and has almost no structural mechanism to say so.

Not the executive with the business case document. Not the board approving the headcount reduction. The engineer who is being asked to review AI outputs she does not fully understand. The analyst whose model is being used to make decisions it was not designed to make. The manager who has been told to roll out a system to her team and who suspects, with some precision, that the efficiency gains her director is celebrating are going to materialise as job losses in twelve months.

These people know something. Their knowledge is specific, grounded, earned through proximity to the actual work in ways that the business case document was not. They are the people the carpenter argument is about. They are the ones who understand whether the understanding is still in the room.

And in most organisations, they have very limited options.

They can raise the concern formally, which in most cultures means being seen as resistant to change, insufficiently enthusiastic about innovation, or, in the specific vocabulary of transformation programmes, a blocker. The professional cost of that label is real. It lands in performance reviews. It shapes how you are perceived in the next restructuring. The culture that says speed is competence also says that the person who hesitates is the person who is falling behind.

They can raise it informally, in the corridor, to a peer who shares the concern. This produces the corridor version of the truth that the meeting version never hears. It is the mechanism by which organisations accumulate large quantities of private correct information that never reaches the decision. Everyone knows. No one has said it in the room where it would change anything.

They can stay quiet and implement. This is what most people do, most of the time, not because they are cowardly but because the structural incentives are consistently aligned against the other options and the personal cost of those options is borne entirely by the individual while the benefit, if the concern is heard and acted on, is distributed across the organisation.

The NHS healthcare workers who refused the Palantir contract are the exception that proves the rule. They had union infrastructure that gave collective weight to individual concerns. They had professional standing that gave their judgment institutional credibility. They had public visibility that made the political cost of ignoring them higher than the political cost of addressing them. Most workers in most organisations have none of these things.

I am not writing this to shame the people who stay quiet. I am writing it because the system that produces the silence is the same system that produces the deployment decisions the silence enables. The executive with the business case is not making a decision in a vacuum. He is making a decision in an environment where the people who could most usefully challenge the framing have been structurally positioned to find the challenge too costly.

That is not an accident. It is the design working as intended.

The leader who wants to break that design does not need a grand gesture. They need one specific, repeated, genuinely held practice. Before any AI deployment decision reaches the point of approval, the people closest to the work must be asked what they see that the business case does not contain. Not as a consultation exercise. As an actual input to the decision. With the standing to change the outcome.

That practice is rarer than it should be. The organisations that have it build differently. The organisations that do not find out what was missing when the system fails.

What augmentation would actually look like

I want to be concrete here because the abstraction makes it easy to miss what is actually being said.

Augmentation, genuine augmentation, has a set of characteristics that are recognisable and measurable. You can check whether it is happening.

The human remains in the decision. Not as a rubber stamp on a machine output. As the actual decision-maker, informed and made more capable by the tool. The surgeon who uses AI assistance to identify candidates for a particular procedure still decides whether the procedure happens. The analyst who uses AI to process the dataset still decides what the analysis means and what should be done with it. Remove the human from the decision and you have crossed the line from tool to replacement.

The productivity gains circulate to the people doing the work. If AI makes a team twice as productive, and the team stays the same size, the humans are working half as much or earning twice as much or some combination. The efficiency dividend does not flow exclusively to the people who own the system. If productivity doubles and headcount halves and wages stay flat, the augmentation framing was a lie. The benefit went to the shareholders. The cost went to the hundred people who lost their jobs and the fifty who remained and are now doing twice the work for the same pay and calling it efficiency.

The human capacity for the work grows, not shrinks. I described earlier the CTO whose team had stopped thinking as hard after five years of AI assistance. The tool had optimised their output while atrophying their judgment. Genuine augmentation does the opposite. The doctor who works with AI diagnostic tools over a decade becomes a better doctor. The engineer who works with AI design assistance over a decade develops a more sophisticated sense of what the tool gets right and wrong and why. The human grows inside the tool relationship, not around it.

The understanding stays in the room. This is what my carpenter was pointing at. When the machine makes decisions, the human loses access to the knowledge of why those decisions are correct. Over time, that knowledge cannot be recovered. When the machine fails, or when the situation falls outside the training data, or when the context has changed in ways the system was not built to anticipate, there is nobody left in the room who knows how to handle it from first principles. The understanding has left. What remains is a room full of people who can operate the machine when it works and are helpless when it does not.

By these four checks, most of what is being deployed under the name of AI augmentation is not augmentation. It is replacement, staged gradually, dressed in the language of tools and assistance and freeing humans for higher-value work. The higher-value work never quite materialises. The lower-value humans are gradually removed. The cycle continues.

The carpenter’s line

Let me go back to the carpenter and the line he drew.

He said, the moment the understanding leaves the room, I am not a carpenter anymore.

He was not talking about job security. He has plenty of work. He was talking about something more fundamental. The relationship between a person and their craft. The knowledge that lives in hands and judgment and years of accumulated experience with the specific behaviour of particular woods in particular conditions. The understanding that cannot be described in a training dataset because it is not declarative. It is procedural, embodied, built into the way his hands move and the way his eyes read the surface of a plank.

A machine that assists him retains his access to that understanding. He uses the tool. He remains the carpenter.

A machine that replaces his judgment removes his access to it. Not immediately. But the muscle that is not used atrophies. The knowledge that is not practiced fades. The understanding that is not applied loses its precision. Within a generation of workers trained to operate the machine rather than understand the wood, the embodied knowledge is gone. Not recoverable from a manual. Not downloadable from a database. Gone.

This is the loss that does not appear in the business case for automation.

The business case shows the cost savings. The cost savings are real. A hundred and forty people costs more than one AI system. The executive with the document was not wrong about the numbers.

What the document does not contain is the accounting for what is lost. The knowledge that leaves the room. The judgment that was built over careers and cannot be reconstructed. The understanding of the actual work, below the interface, that allows a human to handle the situation the system was never trained on.

Every domain of human expertise contains this knowledge. The nurse who knows from the way a patient is breathing that something is changing before any monitor has registered it. The teacher who knows from the quality of silence in a classroom that something happened in the corridor before the lesson. The journalist who knows from the way a source is answering that the source knows more than they are saying.

This is not mysticism. It is pattern recognition of a specific kind. Pattern recognition that is embodied, contextual, and dependent on the human being present in the situation and genuinely responsible for what happens next. Remove the responsibility and you remove the attention that builds the knowledge. Remove the knowledge and you remove the capacity to handle the novel situation.

We are building systems that appear to save cost while actually destroying the knowledge infrastructure that every organisation depends on when the situation is novel. The saving is immediate. The cost is deferred. It will arrive, at scale, when we need people who understand the work and find that we have spent a generation training people to operate systems that do the understanding for them.

The test nobody is running

Here is a question that is almost never asked in the boardroom presentations about AI deployment.

What happens when it fails.

Not fails in the narrow technical sense of a system outage or an error rate above the acceptable threshold. Fails in the deeper sense of encountering a situation that falls outside the training data. A context that has changed. A case that is genuinely novel. The kind of situation that happens in every complex domain with regularity and that requires a human to understand the work at a level below the interface.

I have been in organisations that replaced significant portions of their customer-facing workforce with AI systems and then experienced a crisis. A product recall, a regulatory change, a viral incident that generated an unusual pattern of customer contact at unusual volume with unusual emotional intensity. The AI handled the standard queries. The novel situation, the one that required judgment and empathy and the capacity to say I understand this is not what the script says but here is what I am going to do for you, the system could not navigate.

There were almost no humans left who knew how to navigate it either. Not because the humans who had been displaced were incapable. Because the humans who remained had been operating in a system that handled the judgment calls for three years, and the judgment muscle had atrophied accordingly.

The company managed the crisis. But the cost of managing it, in customer relationships, in regulatory scrutiny, in the emergency retraining of people who had forgotten how to do the thing the system had been doing for them, was not in the original business case. The business case showed the savings from the headcount reduction. It did not show the liability from the capability reduction.

This is a systemic failure of how we evaluate AI deployment. We measure what we can measure. The cost savings are measurable. The knowledge destruction is not, until it manifests as a crisis. And by the time it manifests, the connection between the deployment decision and the capability gap has been buried under years of quarterly reports.

The augmentation version of this story is different. The organisation that deploys AI to assist its workers rather than replace them retains the embodied knowledge. The workers who are made more capable by the tool remain capable when the tool fails. The understanding is still in the room.

This is not a sentimental argument. It is a resilience argument. The organisations that will navigate the crises of the next decade are not the ones that achieved the most aggressive headcount reduction in the previous one. They are the ones that retained the humans who understand the work.

The carpenter has been doing the same work for thirty-one years. He will still be able to do the work if every tool he owns is taken away. That is not inefficiency. That is what decades of genuine expertise looks like.

The person trained to operate the machine that replaced the carpenter cannot do the work when the machine is gone. That is not progress. It is fragility, deferred.

Where this is already happening

I want to name some places where the line is being crossed right now, because abstraction allows the argument to remain comfortable for people who are inside the system making the decisions.

In healthcare, diagnostic AI is being deployed in contexts where the radiologist who used to read the scan is no longer reading it. The AI reads it. A doctor in another country signs off on the output. The local radiologist has been replaced. Not assisted. Replaced. The understanding of the patient’s history, the knowledge of the local disease patterns, the judgment about what a particular anomaly means in the context of this particular person’s previous imaging, that understanding is no longer in the chain. The system is faster and cheaper. When it is wrong, and it is wrong with the specific blindspots of its training data, there is nobody left in the local chain who can identify the error before it becomes a harm.

In journalism, automated content generation is replacing reporters. Not in the narrow technical sense of press release summarisation, which has a reasonable argument for automation. In the sense of local news coverage. The coverage of city council meetings, planning decisions, local court proceedings, the stories that hold local institutions accountable and that require a journalist to be present, to build relationships, to understand the context well enough to know which fact matters and why. This work is being eliminated. Not because AI does it better. Because it is cheaper to not do it. The tool that does not exist is not a more efficient version of the tool that does. It is the absence of the work entirely, disguised as automation.

In education, AI tutoring systems are being deployed as replacements for teaching staff in underfunded districts. Not as assistants to teachers. As substitutes for them. The thirty students in the room are now working with a screen. The teacher who knew which three were falling behind before the test, who knew when the silence in the room was productive and when it was stuck, who knew which student needed to be challenged and which one needed to be left alone today, that person has been replaced by a system that is cheaper and does not require benefits.

The students in those districts are not getting better education with fewer teachers. They are getting education-shaped content delivery without the human relationship that research consistently identifies as the primary predictor of learning outcomes. The children from families that can afford the schools where teachers still exist are not being taught this way. The substitution of AI for teachers is happening where the children of parents with less power have no choice but to accept it.

These are not edge cases. They are the current direction of deployment, in the places where the people affected have the least power to resist it.

The question is whether the people with more power, the leaders inside these organisations, the regulators with the authority to intervene, the workers in adjacent industries who can see the trajectory before it arrives in their own sector, are willing to name what is happening and act accordingly.

The child in the back seat

The carpenter has thirty-one years.

He built the knowledge that lets him know when the machine is wrong over three decades of doing the work, getting it wrong, correcting it, developing the calibration that only accumulates through that specific kind of repeated encounter. The knowledge lives in his hands because his hands have done it thousands of times. The judgment lives in his eye because his eye has learned what the wood does when you get it right and wrong.

The question the essay has not yet asked is what happens to the next generation of carpenters.

Not the people currently in the workforce, who built their knowledge before the systems existed and who can, in principle, maintain that knowledge as long as they keep using it. The people who will enter the workforce in a world where the systems are already present. The child who will grow up in an environment where every diagnostic task, every judgment call, every pattern recognition that takes a practitioner decades to develop, is handled by systems before they have had the chance to build the knowledge themselves.

A radiologist trained on AI-assisted diagnosis learns to evaluate the AI’s outputs. She does not necessarily learn to make the diagnosis. The distinction is invisible in the output. It is everything in the moment the AI encounters a situation outside its training data and the radiologist needs to know what she actually knows.

A journalist trained in an environment where AI tools handle the initial research, the source verification, the pattern identification, learns to work with those outputs. She does not necessarily develop the specific judgment that comes from having done those things badly, been wrong, understood why she was wrong, and rebuilt her approach. That judgment, built through productive failure, is the thing that allows her to notice when the AI is producing plausible nonsense and to know the difference.

This is not an argument against AI in education or in professional training. It is an argument for being honest about what is being traded when we introduce systems that remove the friction of learning. The friction is not an obstacle to the knowledge. In many domains, it is the mechanism of the knowledge. The mistake that hurts, the confusion that resolves slowly, the situation that resists the template these are not inefficiencies in the development of expertise. They are the development of expertise.

A generation of professionals trained to operate systems that handle the difficult parts of their work will be very capable of operating those systems. They will be less capable of the work itself. The gap will be invisible until the systems fail or the situation falls outside what the systems can handle.

That situation arrives in every complex domain. It arrives regularly. It is the nature of complex domains that they produce novel situations the previous framework did not anticipate. The practitioner with thirty-one years of embodied knowledge handles it. The practitioner trained to operate the system that handles it is in a different position.

I have a son. He is eight. He still notices everything the ant carrying something three times its size, the specific way light comes through a particular window at a particular time of year, the sound a door makes that tells him something about the mood of the house. He has not yet learned to outsource his attention. The world is still information that arrives through his senses and requires him to process it.

He will grow up in an environment that is very good at handling that processing for him. Very good at answering his questions before he has had the chance to sit with them. Very good at producing outputs that appear to be the result of the understanding he has not yet built. The environment is not designed against him. It is indifferent to him in the specific way that systems optimised for other things are indifferent to the collateral effects on the people inside them.

The carpenter’s line the moment the understanding leaves the room applies to him as it applies to the radiologist and the journalist and the engineer. The question is whether we are willing to structure his education, his tools, his relationship with the systems around him, in a way that ensures the understanding builds in him rather than being handled by the systems on his behalf.

That is not a technology question. It is a values question. The answer we are currently giving, by default rather than by deliberate choice, is the understanding can be handled by the system. The child does not need to build it.

I do not accept that answer. I do not think we have thought carefully enough about giving it.

What refusing looks like

I want to be specific here because specificity is where most of this argument gets lost in abstraction.

You can refuse replacement-focused AI. Not by rejecting the technology. By insisting on the four characteristics of genuine augmentation and refusing to accept deployments that fail them.

A worker whose role is being automated can ask: am I still in the decision? If the answer is no, that is not augmentation. The company is replacing you, not assisting you. You are entitled to say so. Your union is entitled to say so. Your government is entitled to legislate so.

A government can legislate the productivity circulation. If AI deployment in an enterprise increases output by more than twenty percent, some defined share of that gain goes to the workers whose roles have been transformed. Not as charity. As a legal requirement, on the same basis that minimum wage legislation is a legal requirement. The argument that companies should be allowed to take efficiency gains entirely as profit while workers bear the cost of displacement is a political choice, not an economic law. It can be unmade.

A regulator can mandate the understanding requirement in domains where the understanding matters for safety. Healthcare decisions. Criminal justice. Infrastructure. Education. Financial advice. In each of these domains there are situations where the system will fail or the context will change and a human being will need to understand the work at a level below the interface. Deploying AI in these domains in ways that remove that understanding from the humans in the chain is not a technical efficiency. It is a safety risk, deferred. Regulation can require that the understanding stays in the room.

A society can decide that some domains should not be automated at all. Not because the technology cannot do the task, but because the task requires something that cannot be separated from the human doing it. The nurse’s presence. The teacher’s attention. The elder care worker’s relationship with the person in their care. These are not inefficiencies to be optimised. They are the thing itself. Replacing them with machines does not deliver the service more efficiently. It delivers a different service. A worse one. And it does so while eliminating the livelihoods of people who have built careers in the knowledge that their presence matters.

None of this is technically impossible. The EU AI Act is a beginning. Worker co-ownership models exist. Sector-specific bans exist in some jurisdictions. The robot tax has been proposed and discussed in enough serious policy contexts that it is no longer a fringe idea. The mechanisms are available.

What is missing is the political will, which is a function of power, which is a function of who is in the rooms where decisions are made and what they have agreed to stop accepting.

There is a specific version of this argument that gets made against all of the above. It goes: companies will move their operations to less regulated jurisdictions. Legislating worker protections in one country simply exports the harm to another. International coordination is impossible. Therefore regulation is futile.

This argument is made with great confidence by people who benefit from it being believed.

It is not true. It is a negotiating position. Companies that operate in markets with consumer purchasing power do not, in practice, simply relocate all operations to avoid labour regulation. The history of minimum wage legislation, environmental regulation, and product safety requirements shows consistently that when democracies with significant markets decide something is not acceptable, companies adjust. Slowly, with resistance, with lobbying and legal challenge. But they adjust.

The argument that regulation is futile is itself the primary obstacle to regulation. It is designed to produce the paralysis it predicts. The correct response to it is not to accept its premise but to note whose interest the premise serves.

What this requires, practically, is the same thing every advance in labour rights has required. Workers who are willing to name what is happening. Leaders inside organisations who are willing to say the uncomfortable thing in the room where the decision is being made. Governments that are willing to set the terms rather than wait for the market to arrive at an acceptable outcome on its own. The market will not arrive at an acceptable outcome on its own. It never has. That is what markets are. They are efficient at generating returns for the people who own the capital. They require external constraint to generate acceptable outcomes for the people who do not.

The language we are missing

There is a reason this argument is hard to make inside most organisations. It is not that the people inside them are indifferent to the humans being displaced. Most of them are not. It is that the language available in organisational settings is almost entirely the language of efficiency, cost, output, and return on investment. The language of the other thing, human dignity, the right to meaningful work, the value of understanding that lives in people’s hands and cannot be extracted and replicated, does not have a register in most professional settings.

I watch people who privately hold strong views about what is being lost arrive in meeting rooms and find that they have no vocabulary for the thing they believe. They can speak the language of business risk, of regulatory exposure, of reputational damage if the deployment goes wrong in a visible way. They cannot speak the language of what work means to a person, because that language sounds like sentiment in a context that has defined itself against sentiment.

This is not accidental. It is a managed condition. Organisations that want to make certain decisions without resistance need their decision-making environments to be inhospitable to the language that names what is being decided. Strip the sentiment from the room. Define rigour as the exclusion of the unmeasurable. Train leaders to translate every consideration into a number before bringing it to a table. The result is a professional culture in which the thing that matters most, the human on the other side of the decision, has no language in which to be represented.

Learning to speak that language in professional settings is itself a form of resistance. Not loud resistance. The kind that arrives as a single question in a meeting that has been careful to exclude that kind of question. What does this mean for the people whose roles are affected. Not as a performance of concern. As a genuine ask that requires a genuine answer before the decision proceeds.

The organisations that navigate the next decade well will be the ones that find a way to hold both languages at once. The language of efficiency and the language of the human cost of efficiency. Not because the human cost always outweighs the efficiency gain. Sometimes it does not. But because the decisions made in the absence of that language tend to arrive at places that, on reflection, nobody in the room actually wanted to go.

The executive who said the problem with people is that they have needs was not wrong about the numbers. He was wrong about what the numbers were measuring. He was measuring cost. He was not measuring the knowledge that would leave the room, the resilience that would be lost, the liability that would be deferred, the community that would be damaged, the hundred and forty people who would need to rebuild a working life from the position of having been described as an inefficiency.

Those things are not immeasurable. They are unmeasured. The distinction matters. Unmeasured things can be measured if we decide they are worth measuring. The first step is deciding they are worth measuring. The first step before that is recovering the language that allows us to say why they matter.

The carpenter does not struggle to say why the understanding matters. He has been working in relationship with the material for thirty-one years. The material has taught him. He can read what it needs. He can feel when the tool is right for the task and when a different approach is required. He has the language because he has the experience, and he has the experience because he was never asked to step aside and let a machine accumulate it for him.

That is what augmentation protects. Not the job. The person inside the job. The knowledge that the person carries. The understanding that makes the human in the room genuinely valuable rather than merely present.

There is something that the replacement versus augmentation framing almost captures but does not quite reach.

The framing is still, at its core, an economic argument. We should keep humans in the loop because they provide value that the machine cannot. We should distribute the gains because it is more efficient in the long run. We should maintain the understanding because the organisation will need it when the system fails.

These arguments are true. They are also insufficient.

The reason to refuse replacement-focused AI is not primarily because it is economically suboptimal. The reason to refuse it is that it treats human beings as inputs to a system rather than as the reason the system exists.

The hundred and forty people in the call centre are not cost inefficiencies that the technology has made available to be eliminated. They are people. They have working lives that are a central part of their experience of being alive. They have colleagues, routines, a particular kind of social fabric that forms around work even when the work is not glamorous. They have the knowledge that they are doing something, that their presence contributes to something, that the contribution is recognised and compensated.

When a company eliminates those people and replaces them with a system, it is not just making an economic decision. It is making a decision about what humans are for. It is saying that human beings are valuable precisely and only to the extent that they generate output at an acceptable cost, and that the moment a machine can generate the same output at lower cost, the humans are no longer valuable.

This is a political claim. It is not a technical fact. It is a choice about the purpose of economic organisation. And it is a choice that the people most affected by it, the workers, the communities, the societies that depend on employment as the primary mechanism for distributing participation in economic life, never agreed to and were never asked to agree to.

The question is not whether we can use AI to serve human flourishing. We can. The question is whether we are willing to insist that human flourishing is the point, and that deployments which treat it as a cost to be eliminated rather than a purpose to be served are not progress, regardless of what they do to the profit margin.

The carpenter said it clearly. There is a line. The line is the understanding. When the understanding leaves the room, something has been lost that is not recoverable from a manual.

The companies deploying replacement-focused AI know where that line is. They are crossing it deliberately because the business case says to. The question is whether anyone with the power to stop them will say, with the same clarity that the carpenter said it: this is the line. It does not get crossed.

When your AI tool becomes a weapon

If your team reaches for AI before even trying to understand the problem, your hammer is becoming a weapon.

If your dashboards and AI tools are used more to judge people than to coach them, your system is serving control, not growth.

If only a small group of leaders can change the rules, thresholds, or prompts, your AI is a weapon over the many.

These aren’t glitches in the code they’re design choices about power. Every time you trade transparency for convenience, or autonomy for “efficiency,” you nudge the tool closer to weapon‑territory.

Leaders are accountable for the “power” they hold, are the ones who decide how the hammer is held and who gets to wield it. AI doesn’t just support decisions it reshapes who trusts what, who speaks, and who stays silent. When incentives, metrics, and culture all point toward speed, cost, and compliance, the natural outcome is that every AI tool becomes a weapon of control.

Hammer‑mode leadership does the opposite it builds AI systems that are optional, explainable, and learnable. It rewards people who can still think, debug, and argue without the tool. It treats AI as a training partner, not a verdict machine.

If you can’t imagine your team living without a particular AI, it’s already a weapon over their autonomy. The harder question is what would you need to change so it feels like a hammer again?

What this asks of leaders

I spend my working life in rooms with people who are making these decisions. Not the executives who issue the directives. The people in the middle. The team leads, the engineering managers, the product owners, the architects who are being asked to build the systems that cross the line and who know, in some quiet part of their professional judgment, that the line is being crossed.

Most of them manage the knowledge privately. They build what they are asked to build. They note their reservations in a one-on-one with their manager. They tell themselves that someone above them has assessed the tradeoffs and the decision has been made and it is not their place to refuse.

That is not accurate, it is comfortable.

I want to be honest about why the silence happens. It is not cowardice, exactly. It is something more structural. The person who objects in the room takes on a real professional risk. The person who builds the system and says nothing takes on no immediate risk. The costs of speaking are personal and immediate. The costs of silence are collective and deferred. This asymmetry is by design. It is one of the mechanisms by which organisations produce outcomes that most of the individuals inside them would individually refuse if the choice were put to them directly.

I have sat with engineers who have built targeting systems they did not believe should exist. Customer profiling systems that they knew were being used in ways the data subjects never consented to. Automation roadmaps that they understood would eliminate the roles of people they worked alongside. Every one of them had, at some point, raised a question in a meeting and been told that the decision had already been made at a level above the meeting. Every one of them had accepted this and continued.

The decision at a level above the meeting was made by a person. That person has a name and a salary and the authority to have made a different decision. The acceptance by the people in the room is what gives that decision its operational reality. The system cannot build itself.

This is not a counsel to individual heroism. I am not saying that every engineer should refuse every assignment they have doubts about. The working world does not function that way and the people with mortgages and families and careers cannot be asked to carry the full weight of systemic change as individuals.

What I am saying is narrower. That naming what is happening is available even when refusing is not. That the one sentence, in the room, that describes the actual decision rather than the business case version of it, is always available. That the question this is a replacement not an augmentation, have we assessed the liability of that distinction, is available in every meeting where the distinction matters. That the professional norm of silence, of managing private discomfort while the public decision proceeds, is a norm that can be interrupted without destroying a career.

The person who builds the system that replaces the hundred and forty people is not absolved by the fact that the decision was made above them. They are participating in it. Their technical skills are the instrument of it. Their willingness to execute without naming what they are executing is part of what makes it possible.

Leadership in this context does not require dramatic gestures. It requires saying, in the room where the decision is being made, what the decision actually is. Not the business case version. The version that names the humans being displaced, the understanding being removed from the room, the knowledge being lost, the distinction between the tool and the weapon and which one is being built.

Say it once, clearly and accept that it may not change the outcome. Say it anyway.

Because the alternative, the management of private discomfort while the public decision continues unchanged, is how these things happen without resistance. Not through malice. Through the accumulated silence of people who knew the line was being crossed and decided that naming it was someone else’s job.

The carpenter has been doing the same work for thirty-one years. He has replaced his tools when better ones arrived. He has drawn one line. He draws it not out of sentimentality or fear but out of a precise understanding of what the work is and what would be lost if the understanding left the room.

He is not waiting for someone above him to draw that line.

He is the carpenter. It is his line to draw.

So is yours.

About the Author
Tino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.

When I use the word AI in this essay I am not acknowledging that these systems are intelligent. I am using the word the industry uses, because that is the word that has conquered the conversation. What we are actually talking about is an advanced deep learning model. Extraordinarily capable at pattern recognition, statistics, and probability. Not thinking. Not understanding. Not intelligent in any meaningful sense of the word. There is no academic consensus on what intelligence actually is, and there is certainly no evidence that these models possess it. The word AI is a marketing decision. It was chosen to make the technology feel inevitable, significant, and human. I use it here because refusing to use it would make the essay harder to read. But I want you to know that every time I write AI in this essay, I am describing a very powerful statistical engine. Nothing more. The intelligence is in the room. It is in you.

My son asked me last week what a tree was for.

He is eight. He had been sitting under one for twenty minutes, watching an ant carry something three times its size along a crack in the stone path. He was not bored. He was not seeking stimulation. He was simply present in the way that eight-year-olds are present when nobody has yet taught them that presence is an inefficiency.

I said trees make oxygen. They give us wood. Some trees give us fruit.

He looked at me the way children look at adults who have answered a different question.

What is it for, he said again.

I did not have a better answer. I had descriptions of function. I had economic utility. I had the language of what a tree produces, how a tree performs, what a tree delivers. I did not have an answer to what a tree is for in the way he was asking, which was not a question about output at all. It was a question about meaning. About whether the tree had some irreducible right to exist that had nothing to do with what it gave us.

I sat down on the path next to him and watched the ant.

We stayed there for a while. Neither of us said anything useful.

The question under every question

I have spent the last several years writing about AI, about power, about what technology is costing us as humans. I have written about the system that made work the answer to identity. I have written about the bargain that is breaking. I have written about the designer who built the beautiful interface without understanding what it was doing, and about the Mediterranean cultures that kept alive a way of being in the world that does not depend on employment to feel whole.

All of those essays were, in some sense, asking the same question. They just never quite asked it directly.

The question is this.

Do we actually want what we have been building toward.

Not do we want AI. Not do we want automation. Not even do we want efficiency or productivity or the particular kind of progress that the last two centuries have been organised around.

Do we want the destination. The thing at the end of the road we have been on. The world that arrives if the logic we have been following is followed all the way.

I have been sitting with this question for a long time. I find it uncomfortable in a specific way. Not the discomfort of a question that is difficult to answer. The discomfort of a question that I suspect I already know the answer to and am not ready to say it out loud.

Because if the answer is no, then the problem is not how we regulate AI. It is not how we tax the offshore wealth or redistribute the efficiency gains or retrain the displaced workers. Those are all important questions, urgent questions, questions worth spending careers on. But they are downstream of a more fundamental choice about what kind of creature we are and what kind of world we are trying to build.

And that question is not a policy question. It is a philosophical one. A cultural one. Possibly a spiritual one, if that word can survive in a conversation about technology without becoming its own kind of evasion.

Precision as religion

A knife I have had for eleven years lives in the third drawer of my kitchen. It was given to me by a chef I once worked alongside briefly, in a different life, before the tech career consumed everything. He told me when he handed it over that a good knife is not precise. It is responsive. A precise knife does the same thing every time. A responsive knife does what the material needs. The difference, he said, is everything.

I have thought about that distinction more in the last two years than I thought about it in the previous nine.

Precision is the god of the current age. We worship it openly and without embarrassment. We build systems to eliminate the human variability that introduces imprecision. We measure everything that can be measured so that we can optimise everything that can be optimised. We have constructed an entire civilisation around the premise that precision is inherently better than its absence. That the precise answer is more valuable than the responsive one. That the consistent output is worth more than the one shaped by the moment and the material.

AI is the apotheosis of this religion. It is the most precise instrument we have ever built. It does the same thing, more or less, given the same input. It does not have bad days. It does not bring the residue of a difficult conversation into the next one. It does not read the room and decide that what the room actually needs is different from what it was asked for.

We look at this and call it intelligence.

I want to ask what we mean by that word. Not to score a philosophical point but because the confusion is consequential. When we call AI intelligent, we are making a claim about what intelligence is. We are saying that intelligence is pattern recognition at scale. That it is the capacity to retrieve and recombine information faster and more accurately than any human could. That it is, in short, a form of precision applied to cognition.

But every teacher I admired in my life was not precise. They were responsive. They could read a room of thirty students and feel which three were lost and which one was bored and which one needed to be challenged and which one needed to be left alone today because something was happening at home. They could not have told you how they knew. The knowing was in their hands and their eyes and their thirty years of being in rooms with children. It was imprecise, uncodifiable, irreducible to an algorithm. And it was, by any meaningful measure, intelligent in a way that no pattern-recognition system has yet approached.

The precision religion cannot account for that kind of intelligence because it cannot measure it. And what it cannot measure, it tends eventually to dismiss.

What we were running from

There is a story I want to tell about a man I met at a conference in Lisbon three years ago.

He was a founder. Mid-forties. His company had just been acquired for a number that made him, by any conventional measure, secure for several lifetimes. He was at the conference not because he needed to be but because he did not know what to do with a Tuesday that did not have a schedule.

We ended up talking for two hours at the edge of a rooftop bar, the city orange in the early evening, the smell of salt coming up from the river below. He was successful by every metric the productivity culture recognises. He was also, very quietly, one of the most lost people I had met in years.

He said: I thought the acquisition would feel like something. Like I had arrived somewhere. I keep waiting for the feeling.

I asked what feeling he was waiting for.

He thought about it. Then he said: that it was worth it.

He was not talking about money. He had the money. He was asking whether the thirty-hour days and the missed dinners and the relationships that had not survived the velocity and the identity so completely organised around his company that when the company was sold he genuinely did not know what he was anymore, whether all of that had been in the service of something. Whether there was a destination that justified the road.

He had achieved everything the system promised achievement looked like. And he was standing on a rooftop in Lisbon asking a stranger whether any of it had been worth it.

This is not an unusual story. I hear versions of it regularly. What is unusual is that he was willing to say it out loud, in those words, without the protective layer of lessons learned or pivots to the next chapter that most successful people deploy when the conversation gets close to the actual question.

The actual question, underneath his question, is the same one my son was asking about the tree.

Not what does it produce. What is it for.

We have built an entire civilisation of production without ever properly asking that question. Or rather, we asked it and then accepted an answer that turned out to be, on close examination, a circular reference. Work is for productivity. Productivity is for growth. Growth is for prosperity. Prosperity is for the good life. The good life is for, roughly speaking, more of the same.

The man on the rooftop had followed that logic all the way to its conclusion and found it empty.

He is not alone. He is just unusually honest.

The delusional ape

I want to use a phrase that was offered to me recently in a conversation about AI and identity. The person I was talking with, frustrated with the circularity of most AI discourse, said this.

“We are a delusional ape hallucinating narratives as we traverse this reality.”

It is not a kind description. It is also, I think, more true than most descriptions we use.

While some believe we are biological creatures who arrived through an unintentional process, there are those that believe that we are human beings who came for a purpose and are heading toward a defined destination. We developed the capacity for consciousness, which is extraordinary and still largely unexplained. And we used that consciousness primarily to construct stories about why we are here and what we are supposed to be doing and whether we are doing it correctly.

Every culture in human history has done this. The stories differ. The need to have a story does not.

The productivity gospel is one of these stories. It says: the purpose of a human life is to produce value. Progress means producing more value more efficiently. The good society is one organised to maximise production. The good life is one that contributes maximally to that production.

This story has been extraordinarily successful at generating material wealth. It has also been extraordinarily successful at generating misery of a particular kind. The misery of people who have followed the story faithfully and arrived at its promised destination and found it does not feel like destination at all. Who have produced and produced and optimised and achieved and looked up from the spreadsheet of their life to find the rooftop in Lisbon and the question that has no answer in the story’s own terms.

What the AI era is doing, among other things, is stress-testing this story at scale. If the purpose of a human life is to produce value, and machines can produce that value more efficiently, then what is the purpose of a human life.

The story cannot answer that question. Because the story was never really about purpose. It was about distraction. A narrative complex enough to occupy the delusional ape’s extraordinary consciousness so thoroughly that the underlying questions, what are we for, what do we owe each other, what does it mean to live rather than merely function, could be safely deferred.

AI is removing the distraction. Not intentionally. Not kindly. But unavoidably.

What is left when the distraction is gone is the question my son was asking about the tree.

What machines cannot know

I want to be careful here because this is where the argument is tempting to get wrong.

The wrong version goes: humans are special, AI cannot replicate consciousness, therefore AI is not really intelligent, therefore the threat is overstated.

That is not what I am saying.

What I am saying is something different. That there is a category of knowledge that requires being mortal, embodied, and uncertain to access. And that this category of knowledge is not a small addendum to human intelligence. It is its foundation.

A surgeon who has never been afraid does not understand what it costs a patient to put their body in another person’s hands. A leader who has never failed does not understand what it takes for a team member to admit they are struggling. A parent who has never lost something essential does not know the particular quality of attention you give to what you have and are afraid of losing.

This is not sentimentality. It is epistemology. The knowledge that comes from being vulnerable, from having stakes, from being the kind of creature that can lose things and be changed by losing them, is a different category of knowledge from the knowledge that comes from pattern recognition across large datasets.

I think about this when I watch leaders try to use AI as a substitute for the difficult conversation. The conversation where someone needs to be told that what they are doing is not working. Where someone needs to hear, from a person they trust, that the direction is wrong. Where someone needs the specific experience of being in the room with another human being who has assessed the situation and is now offering them the honest version rather than the diplomatic one.

AI can generate the words. It can produce language that is, in many cases, more precise than what the leader would have said unprompted. More structured. Less emotional. Less contaminated by the relationship between the people in the room.

But the contamination is the point.

The reason the conversation matters is because it happens between people who have stakes. Who have history. Who will still be in the room together next week and the week after. Who are changed by what is said and by the fact of having said it. The weight of the conversation is not a problem to be designed out. It is the mechanism by which the message lands differently than it would on paper.

A piece of feedback delivered by someone who cares about you and is afraid of losing your respect and is choosing to say the difficult thing anyway is not the same piece of feedback delivered as a well-structured paragraph generated by a system that has no relationship with you and cannot lose anything by saying it.

The words might be identical. The knowledge transmitted is entirely different.

AI will get better at simulating the language of that knowledge. It will produce outputs that look, in many contexts, indistinguishable from the real thing. This is already true and will become more true.

But simulation is not the thing. A map of the territory is not the territory. The question what does it mean to grieve is not the same question whether you can answer it or not.

The danger is not that we will mistake AI for human. The danger is that we will mistake the simulation for sufficient. That we will accept the map because the territory is too difficult to navigate. That we will build systems around the convincing imitation of human understanding and call that understanding, the way we built systems around the convincing imitation of human judgment and called that efficiency.

The chef who gave me the knife was teaching me something about the difference between responsive and precise. The responsive thing is harder. It requires presence, attention, accumulated knowledge that lives in the hands as much as the head. It cannot be replicated by a system that has never held the knife.

We are building a world in which the precise thing is systematically preferred over the responsive thing. Because the precise thing is cheaper to scale. Because the responsive thing requires the kind of human presence that cannot be extracted and replicated. Because the responsive thing keeps the human in the room and the human is, from a certain angle, the most expensive component in the system.

That preference is a choice. It is also, I want to argue, a mistake. Not just ethically. Practically.

The map and the territory

I talked to a CTO last month who had just finished a six-month implementation of an AI decision-support system for his team. The system was excellent. It was fast, it was consistent, it was right more often than the humans it was designed to support. By every metric the project had been measured against, it was a success.

He called it a success with the particular flatness of someone describing something that has cost more than the balance sheet shows.

I asked what was not on the metrics.

He was quiet for a moment. Then he said: my best people have stopped thinking as hard.

He did not mean they were lazy. He meant that the presence of a system that produced correct answers reliably had altered the relationship between his team and the problems they were solving. They were now problem-checkers more than problem-solvers. They reviewed the system’s output rather than generating their own. The cognitive muscle that gets built by sitting with a difficult question and not knowing the answer and having to find your way through it was being exercised less, and atrophying accordingly.

The system was optimising their output while quietly eroding their capacity.

This is not a new phenomenon. It is what calculators did to mental arithmetic. What GPS did to spatial reasoning. What spell-check did to the intuitive understanding of how words are constructed. Each individual substitution looked like efficiency. The cumulative effect was a slow narrowing of the range of things the human could do without the tool.

But mental arithmetic and spatial reasoning and spelling are relatively small losses. What the CTO was describing was a loss at a different order of magnitude. The capacity to think hard about difficult problems is not a peripheral skill. It is, arguably, the central skill. The thing that makes a team capable of navigating a situation the system was never trained on.

The territory of reality is not the map of the problems the system has seen before. Novel situations arrive. Situations the training data never contained. Situations that require a human being to sit with uncertainty and confusion and incomplete information and still make a judgment.

If the humans who are supposed to make those judgments have spent five years checking the system’s output instead of developing their own, they will not be ready when the situation arrives that the system cannot handle.

We are optimising for a world that does not exist, the one where the system always has an answer, while eroding our capacity to navigate the world that does, the one where sometimes there is no answer and you have to make your best guess from the position of a mortal creature with incomplete information and genuine stakes in the outcome.

The civilisation that forgot it was alive

The mechanistic worldview has a logic and the logic is internally consistent. Reality is a system. Systems can be understood. Understanding can be translated into control. Control can be translated into optimisation. Optimisation is progress.

This logic has produced antibiotics and clean water and agricultural yields that feed populations that would have been unimaginable two centuries ago. I do not want to romanticise a past that involved dying of infections and watching children starve. The mechanistic worldview has delivered real things of real value and anyone who dismisses it entirely is doing something dishonest.

But.

The mechanistic worldview is a tool, not a truth. It is a way of modelling reality that is extremely useful for certain kinds of problems and useless for others. The problem is that we have elevated it from tool to worldview. From a method to an identity. We have organised societies, economies, institutions, and finally the way we understand ourselves around a model of reality that was built for the manipulation of physical systems and applied, wholesale, to the question of how to live.

The result is what you see on the rooftop in Lisbon. A man who has succeeded by every measure the system offers, standing in the orange evening light, waiting for a feeling that the system was never designed to deliver.

The system optimises for measurable outputs. Meaning is not a measurable output. Love is not a measurable output. The particular quality of attention you give to a eight-year-old sitting under a tree watching an ant is not a measurable output. Beauty is not a measurable output. The kind of trust that is built between people over years of difficulty shared and survived is not a measurable output.

These things are not inefficiencies. They are what the efficiency is supposed to be for. And we have built a civilisation so thoroughly organised around the measurable that we have systematically devalued everything that falls outside the measurement.

What interests me is how thoroughly this worldview has colonised even the vocabulary available to us when we try to resist it. We cannot argue for the value of the three-hour meal without framing it in terms of wellbeing metrics and productivity benefits and research showing that social cohesion correlates with economic resilience. We cannot argue for rest without citing studies about cognitive performance after recovery. We cannot argue for the unmeasurable without first translating it into the measurable, because the measurable is the only language that the current system accepts as legitimate.

This is the deepest form of the problem. Not that we value the wrong things. But that we have lost access to a language for valuing things that cannot be measured. The language of the intrinsic. The language that says: this is worth something because it is, not because of what it produces.

My son has that language. He is eight. He has not yet traded it in for the other one.

The cultures that maintained the three-hour meal also maintained that language. Not as a luxury or a philosophical indulgence. As a practical necessity. As the thing that allowed them to keep building communities where people wanted to be rather than places people had to be because the economic logic left them no other option.

AI is arriving into a world where that language is already endangered. Where the people who might resist the substitution of the simulated for the real are working in the vocabulary of the system they are trying to resist. Where the argument for human connection gets made by citing engagement data and the argument for rest gets made by citing productivity research and the argument for the unmeasurable gets abandoned because nobody in the room where decisions are made has the language for it anymore.

This is the loss that is hardest to name. And the hardest to recover from.

What the Mediterranean kept

The cultures that economists spent decades describing as inefficient were, among other things, holding something that was not on any balance sheet.

A meal in a village in southern Portugal takes three hours. Not because the food takes three hours. Because the meal is not primarily about the food. It is about the particular alchemy that happens between people when they sit together without agenda and let the conversation go where it goes and allow the time to be what it is rather than what it can be extracted into.

This is not nostalgia. I am aware that those same villages contain their own forms of constraint and cruelty, their own hierarchies and injustices, their own things that should not be preserved. I am not arguing for a return to an imagined pastoral past.

I am arguing that the three-hour meal contains a piece of knowledge about what it means to be human that the productivity gospel cannot encode. That the thing happening in that meal, the thing that is not the food, is not a luxury or a bonus or a cultural affectation. It is a fundamental human activity. The building and maintenance of the bonds that make a person feel they exist in a world rather than merely passing through it.

The knowledge that this contains is not abstract. It is practical. Societies that maintained these structures retained a social fabric that, when economic catastrophe arrived, gave people something to stand on that was not their job title or their salary or their professional identity. They had each other. Not metaphorically. Concretely. People who would bring food. People who would sit with you. People for whom your value was not conditional on your output.

This is what the mechanistic worldview cannot produce, not because it cannot value these things in principle, but because it has no mechanism for measuring them and therefore no mechanism for protecting them when they conflict with something that can be measured.

We are now in a moment where what can be measured is being automated and what cannot be measured is being revealed as the only remaining irreducibly human territory. The delusional ape, it turns out, cannot be optimised out of its need for exactly the things that optimisation cannot account for.

And here is what is interesting about that. The AI companies understand this, at some level. The language they use to sell their products is saturated with the vocabulary of human connection. More time for the things that matter. Focus on what only you can do. Get back to the work that is truly yours. The promise is not efficiency for its own sake. The promise is that efficiency is the route back to humanity.

This is the most sophisticated version of the problem. Not that we are being sold a machine and told it is a tool. But that we are being sold efficiency and told it is the road to meaning. That by automating the tedious, we will recover the time and the energy to do the genuinely human things. That the machine is not replacing the human. It is freeing the human to be more fully human.

It is a beautiful story. It might even be true for some people in some contexts. The problem is the track record. The agricultural revolution was supposed to free humans from backbreaking physical labour. It did, for some people, in some places. It also created the conditions for industrial capitalism, which created new forms of backbreaking labour and new systems for extracting human time on behalf of people who owned the machines. The story the machines tell about themselves tends to be more optimistic than the story the people inside the system experience.

The question is not whether AI will free some time for some people. It will. The question is whose time, and freed for what, and who decides, and what happens to the people whose time is not freed but simply made redundant.

Those are not questions the technology answers. They are questions the politics decides. And the politics, right now, is being run by people who have a significant financial interest in a particular set of answers.

Do we want this

I want to return to the question I have been circling.

Do we actually want what we have been building toward.

I do not think most people do. I think most people, if you sat with them long enough on a rooftop in Lisbon or at a kitchen table in the afternoon or under a tree in a garden while a child watched an ant, would tell you that what they want is not more efficiency or more precision or more optimised engagement with algorithmically curated content.

They want to feel that they are here. That the people they love are here. That the work they do means something beyond its contribution to a metric. That they are participating in a life rather than executing a function.

These are not complicated wants. They are the wants of the delusional ape. The wants that every culture in human history has tried to address through ritual and religion and community and art and the particular way that humans have always organised themselves around fire and food and the stories that make the darkness less absolute.

AI is not answering those wants. In many cases it is specifically designed to redirect them. The engagement economy that funds most AI development is built on the premise that human wants can be substituted with simulations. That the want for connection can be satisfied with a notification. That the want for meaning can be addressed with a personalised content feed. That the want for recognition can be met with a like, which costs the giver nothing and delivers the receiver a small neurological reward that wears off in minutes and requires replenishment.

This is not a conspiracy. It is a business model. And it has been extraordinarily successful by the measures the business model uses, which do not include whether the humans inside the system are actually getting what they want.

I watched a senior engineer at a company I was advising try to explain to his teenage daughter what he had been working on for the previous three years. She was sixteen. She listened politely. Then she asked: but what does it do for people.

He described the product. The engagement metrics. The daily active users. The revenue model.

She said: but what does it do for people.

He understood what she was asking. He did not have an answer that satisfied either of them. Because the answer was honest and the honest answer was that it did something to people rather than for them. That the engagement metric and the human good were not the same number, and the company had spent three years optimising for the number while the other thing had been left to look after itself.

The sixteen-year-old was asking the same question my son asks about the tree. Not what does it produce. What is it for.

The question is not new. Every generation asks some version of it. What is different now is the scale and the speed. We have built systems of extraordinary reach, operating at a speed that leaves no room for the question to be asked between iterations, optimising for measures that were chosen for their measurability rather than their alignment with what actually makes a human life feel worth living.

And those systems are now building the next generation of systems in their own image.

The question do we want this is not a question about technology. It is a question about whether we are willing to look clearly at what we have been choosing and decide whether to keep choosing it.

The choosing is real. This is not inevitable. The three-hour meal did not disappear because of some unstoppable force of human nature. It disappeared because a set of economic and cultural choices systematically devalued the things it produced. Those choices can be made differently.

The question is whether we are willing to make them. Which requires first being willing to name what we actually want, as opposed to what the system has trained us to say we want.

There is a version of this that becomes a counsel of despair. A litany of losses that ends with the suggestion that modernity was a mistake and we should go back to some simpler life that never actually existed in the form we imagine it. That is not what I am saying and I want to be clear about that.

What I am saying is narrower and more practical. That the tools we build should serve the wants of the creature using them, not reshape the creature into something that fits the tool better. That when we notice a gap between what the metrics measure and what actually matters, the response should be to question the metric rather than to dismiss what the metric cannot capture. That the three-hour meal and the knife that is responsive rather than precise and the eight-year-old watching the ant are not romantic anachronisms. They are data about what human beings actually need that our measurement systems are not equipped to collect.

My son under the tree was not confused about what he wanted. He wanted to understand what the tree was for in a way that had nothing to do with output. He wanted the world to make sense at a level below function and utility and measurement.

He is eight. He has not yet been taught to call that wanting impractical.

I spent fifteen years in organisations that would have called it impractical. I spent a Saturday morning at a kitchen table noticing that I could not sit still with nothing to show for the hour, and understanding that the inability was not mine but the system’s, installed in me over years of being inside it.

I am still working on the uninstalling.

The delusional ape hallucinating narratives as it traverses this reality. That is what we are. Precisely and completely. Not as an insult. As a description of something extraordinary. We are the universe looking at itself and making stories about what it sees. We are the only thing we know of that does this. It is not a bug. It is the whole point.

The question is which stories we choose to tell.

The story we have been telling, that precision is progress, that output is value, that the optimised life is the good life, is running into its own limits. The arrival of AI is not creating those limits. It is revealing them. The story always had a problem at its core. The problem is that it described a destination nobody actually wanted to arrive at.

The man on the rooftop arrived at it. He was waiting for a feeling the story had no mechanism to deliver.

My son under the tree was asking for the thing the story had no language to describe.

The tree does not produce anything right now. It is simply there. Existing completely. Doing the thing that it is, without justification or apology or quarterly metrics.

There is a kind of knowledge in that. The kind that lives in hands and eyes and thirty years of being in rooms with things you have paid attention to. The kind that comes from being mortal and present and genuinely uncertain about what happens next.

The knife is responsive, not precise.

That is the difference. That is everything.

If this essay landed, the Tuesday posts come by email. Free. Subscribe below and the next one arrives in your inbox.

If you are ready for the room, paid subscribers get deeper essays, the book as it is being written, monthly live sessions, and direct access to the thinking before it is polished. The door is open.

From this series;

1 - https://newsletter.diamantinoalmeida.com/p/who-are-you-without-the-title

2 - https://newsletter.diamantinoalmeida.com/p/they-did-not-accidentally-make-work

3 - https://newsletter.diamantinoalmeida.com/p/a-delusional-ape-hallucinating-narratives

4 - https://diamantinoalmeida.substack.com/p/the-hammer-and-the-weapon

About the Author
Tino Almeida is a tech leader, coach, and writer reshaping how we think about leadership in a burnout-driven world. With over 20 years at the intersection of engineering, DevOps, and team culture, he helps humans lead consciously from the inside out. When he’s not challenging outdated norms, he’s plotting how to make work more human, one verb at a time.