What Will AI Do to Us? It Will Turn Every Pro Into an Amateur.

The End of Productive Activity

What will AI do to us?

People answer with a list. It will write software. Discover medicines. Prove theorems. Run companies. Replace junior engineers, then senior engineers, then managers, then—perhaps, someday—founders.

I think that misses the point.

AI will turn every pro into an amateur.

And here is the trap: if your strategy for surviving AI is to become more productive with AI, you're optimizing the very thing AI is making obsolete.

TL;DR

Most people are hired because they can produce an outcome better, faster, or cheaper than the customer can. That is a productivity relationship. When sufficiently capable AI lets customers obtain that outcome directly, the reason to hire the human disappears. Becoming the employee who uses AI most efficiently can help you today—but it is not a permanent defense. You are helping demonstrate that the output can be separated from the person.

Chess is not a counterexample: professional chess survives superhuman engines because its audience buys human rivalry and entertainment, not the mathematically strongest move. Even that emotional market is not guaranteed to remain immune to AI.

The endgame is not a world where the most productive human wins. It is a world in which productive activity no longer needs a human.

Not because you will forget how to do your work. Not because you will stop loving it. But because the world will no longer need to pay you to produce what intelligence can produce on demand.

You can still be a mathematician. You can still write software. You can still analyze stocks, make films, tell jokes, and build things.

The activity can remain. The profession may not.

And if that sounds implausible, start with a question so ordinary we rarely ask it.

Why does anyone pay anyone?

You want something you cannot make yourself.

You need software, so you hire a programmer. You need a market analyzed, so you hire a consultant. You need a campaign, so you hire a marketer.

In each case, there is a gap between your desire and your ability to satisfy it. A business gets paid to close that gap.

Desire → inability to satisfy it yourself → paid specialist → result.

Now put sufficiently capable AI into the middle.

You describe the software you want. It builds, tests, deploys, and maintains it. You ask for an investment analysis. It gathers evidence, models possibilities, and explains risks. You want a campaign. It designs and runs one.

Desire → your AI → result.

The striking part is not that a machine has taken the specialist's seat. It is that the buyer may no longer need to hire a specialist at all.

This is what makes AGI different from another productivity tool. The tool does not simply help the professional do the job. It lets the customer do without the professional.

You can still be a mathematician

Imagine living alone in a treehouse in Norway.

No internet. A shelf of mathematics books. A notebook. Years of uninterrupted thought.

After ten years, you discover a beautiful theorem. You have experienced the pleasure of understanding something for yourself. That experience is real.

Then you reconnect and discover that AI proved your theorem years ago—and extended it in directions you never imagined.

Was your decade wasted? Not if you loved mathematics.

But why would someone pay for your result when the result is already available?

You can keep doing mathematics. You just cannot assume that doing mathematics will remain an economically scarce service.

This is what I mean by amateur. Not an unskilled person. Someone who does an activity without needing a customer to fund it.

A programmer can program for pleasure. A researcher can investigate out of curiosity. A writer can write because they want to write.

AGI does not have to end the activity to end the profession.

The end of productive activity

Here is the deeper implication.

This is the end of productive activity.

By that I mean activity whose primary justification is an external result: more output, more efficiency, more money, better decisions, better performance.

You learn a skill to become employable. You write code to ship a product. You analyze a company to make a better investment. You optimize your day to get more done.

When AI can produce the result more effectively, the productivity argument for doing it yourself collapses.

This is a thesis about sufficiently capable, widely accessible AI—not a claim that all economic work has already disappeared. Physical resources, institutions, ownership, and access still matter. But none of those guarantees that human productivity must remain the point of human life.

We have been trained to ask: What did you accomplish? What did you produce? How useful were you?

What if those are no longer the important questions?

The most dangerous career advice in the AI era

You've heard it everywhere: AI won't replace you. Someone using AI will.

For the present transition, that can be practical advice. An engineer using AI may outperform an engineer who refuses it. A marketer using AI may produce more campaigns. A consultant using AI may serve more clients.

But look at the direction of travel.

First, one employee with AI does the work of five. The company celebrates. Then one employee with better agents does the work of fifty. Then the customer uses those agents directly.

The productivity improvement you proudly demonstrated becomes evidence that the company needs fewer people—including you.

This is the paradox: you are training yourself to compete on the dimension where AI's advantage is growing fastest.

Silicon Valley has built an entire culture around meritocracy: reward the people who ship faster, solve harder problems, and produce more value. The best engineer, the most effective operator, the founder who executes relentlessly—these are the people we celebrate. And for good reason: in today's economy, results matter. But that is precisely the trap. If the work itself can be performed autonomously, being the most productive human does not guarantee that a human is needed at all.

Imagine saying, “I'll keep my job because I'm better at using AI than my colleagues.” What happens when the AI no longer needs you to operate it?

You may win the competition against your coworkers and lose the competition against the need for a coworker at all.

Using AI to increase productivity is a strategy for the transition. It is not a strategy for surviving the destination.

The twenty-second illusion

People like to rank jobs by how long they think they will survive.

Junior engineer first. Senior engineer later. Manager after that. CEO much later. Founder last.

Everyone above the next rung gets a comforting story: My work is different. My judgment is special. I still have time.

But recursive self-improvement, if it becomes fast and effective, challenges the idea of a leisurely ladder.

Imagine the capability gap from junior engineer to founder collapsing in twenty seconds.

Twenty seconds is a thought experiment, not a forecast. Deployment, regulation, capital, and institutions may make the real transition much slower. But the thought experiment exposes a fragile assumption: that being harder to automate grants you decades of meaning.

Suppose you celebrate because AI can replace every employee but still needs you, the founder.

Twenty seconds later, it no longer does.

What exactly did you win?

Do you want to build your identity around being the last person whose job becomes unnecessary?

Why does Magnus Carlsen still get paid?

AI-generated illustration; not a documentary photograph of Magnus Carlsen.

Here is the strongest objection.

Chess engines surpassed human players long ago. Yet people still watch Magnus Carlsen. They still care who wins. Sponsors still support tournaments. Human players still have careers.

Doesn't this prove professionals can survive superhuman AI?

No. It proves something more interesting.

Imagine a chess game in which losing means you die. You can choose a top engine or a human grandmaster to play for you.

You choose the engine. You want the strongest moves.

Now imagine watching Magnus face a rival in a championship. You do not tune in for the objectively strongest moves possible. You tune in because Magnus is playing. There is a history, a rivalry, a human being under pressure, a possibility of triumph or collapse.

The engine sells strength. The human match sells a story.

The audience is not buying the best move.

A human chess tournament is closer to live sport than to consulting. Identity, presence, constraint, and narrative are part of the experience.

Imagine hundreds of competitors with devices that let them play at superhuman strength. Their moves might be extraordinary. But their tournament would not automatically be more compelling than the human rivalry people already care about.

More skill is not the same thing as more desire to watch.

Chess is entertainment, not a productivity service

The crucial distinction is the business you are actually in.

A professional chess player looks, superficially, like a knowledge worker: sitting at a board, calculating moves, using extraordinary intelligence. But the paying audience is not purchasing the most accurate calculation. It is purchasing entertainment, suspense, identity, and the unfolding story of particular people.

A tournament of the strongest chess agents might be fascinating to researchers, engine developers, or benchmark enthusiasts. It would not automatically command the same emotional audience as a match between recognizable human rivals. That is not a claim that nobody watches computer chess; it is a claim that stronger play alone does not create the same product.

Now consider a bakery. Most customers want bread that tastes good, is convenient, and is worth the price. If autonomous systems can consistently make better bread more cheaply, the ordinary bakery competes on those outcomes. A bakery might also sell a ritual, a neighborhood identity, or affection for a particular baker—but those are distinct reasons to choose it, and a good story does not rescue bad bread for customers primarily buying bread.

This applies far beyond office work. The test is not whether the worker uses their hands or their brain. The test is what the buyer is paying for.

Is it the result? Or is the experience of this particular person doing it part of the result?

Your consultancy is not a chess tournament

When a company hires a consultant, it usually wants a result.

Lower costs. Better strategy. Higher revenue. A problem solved.

It is not paying to watch the consultant struggle heroically with a spreadsheet.

When a customer buys SaaS, they usually want software that works. The founder's biography might make a great podcast. It does not make a broken product useful.

When people switch between AI assistants, they reveal the same preference. Many move toward whichever system better serves their needs. They are not necessarily buying loyalty to Sam Altman, Dario Amodei, or a corporate mission.

Of course people care about reliability, privacy, habit, price, and trust. But these are often part of how effectively the product serves them.

A founder can be charismatic and still lose customers. A consultant can be beloved by a team and still be unnecessary to the buyer.

You can win the internal status game and lose the external market.

Even the founder

Founders often assume they sit outside the automation story.

They say: AI can write code, but it cannot decide what to build.

Why not?

Identifying needs, generating hypotheses, designing products, running experiments, allocating budgets, and changing strategy are all activities an increasingly capable agent could undertake.

They say: AI can do the work, but the founder owns the company.

Ownership is not the same as being necessary to operate it.

They say: AI can build a product, but someone still has to sell it.

What if the customer no longer needs to buy the product because their own AI can make an equivalent solution?

The founder is not the final boss of automation.

The founder is another participant in the economy of satisfying desires. And if that economy changes, the founder changes with it.

Fine. Then sell a story.

This is where people retreat to storytelling.

Build a personal brand. Create a community. Make customers emotionally attached. Become someone they want to support.

There is something real here. People pay for performances, fandom, identity, belonging, and particular human relationships.

But do not confuse a story attached to a product with a story that is the product.

Imagine a famous creator with a million followers launching an ordinary B2B analytics tool. Businesses buying analytics still want accurate answers, integrations, dependable operations, and good economics.

The creator's story may bring attention. It does not remove the benchmark.

Or consider a data science course. Surround it with inspiring stories about becoming a creator or finding yourself. But if students mainly want to learn data science, they will compare it with an AI tutor that teaches them personally and instantly.

The same challenge applies to an affirmation app.

If you sell affirmations as a way to feel better, become confident, or change behavior, you promise an outcome. An AI that knows the user better and adapts in real time can compete for that outcome.

What if the app instead contains a creator's daily reflections, and people subscribe because they want to follow that particular life?

Now the product is closer to a relationship or an ongoing performance.

That is different. But it is not invulnerable.

AI can tell stories too

Suppose AI can generate characters, relationships, communities, voices, and histories more compelling than most human creators can.

Suppose it tailors a story to your memories, humor, fears, and interests. Suppose robots and immersive environments eventually extend those experiences into physical space.

Why assume storytelling is an exclusively human profession?

It isn't enough to say, “But people will know it is really me.”

Some will care deeply. Others will care more about the experience. A real relationship may remain valuable without guaranteeing a scalable paid profession.

There is no magical product category that AI cannot compete with simply because we call it human.

Even community, emotional connection, and authenticity must face the same question:

What desire is being satisfied? Why does this person need to pay you to satisfy it?

If the answer is only that you currently provide the best experience, a better experience can replace you.

Distribution is not just getting attention

For builders, this changes the meaning of distribution.

A feature comparison can persuade someone to try your software. A benchmark can show it performs better. But if that is the entire reason people stay, another model can take them away as soon as it wins the benchmark.

A story is not a slogan on an About page. It has to shape what people actually come to experience.

For Manager, this is a strategic challenge, not an exemption from our thesis. If Manager promises only more effective behavioral interventions, a more capable AI can compete on intervention quality. If it promises only a faster way to accomplish goals, users can compare it on speed. Distribution cannot magically transform an outcome-based product into an irreplaceable relationship.

What we can do is make our purpose legible through concrete work: publicly document the questions we investigate, show what we build and test, disclose failures, explain decisions, and let people follow an ongoing intellectual and human journey rather than only a feature list. The product must still work. The story must be real. And neither is a guarantee of permanent immunity.

A strong story may win attention and loyalty when functional performance converges. It is not a substitute for performance, and AI may eventually compete for that loyalty too.

What should you build?

Stop searching for a permanently AI-proof business.

You can build products today. You can make money today. You can serve customers today. Be clear about what people buy, and why.

But do not confuse a current opportunity with a permanent human advantage.

If your product sells an outcome, expect competition on that outcome.

If it sells an experience, ask whether AI can create a more desirable experience.

If it sells access to you, ask whether the audience genuinely values you rather than the utility you provide.

And do not turn this into another optimization checklist for staying economically relevant forever. That misses the point.

The more important question is what you do when relevance is no longer something you can win through productivity.

What would you do for nothing?

For me, one answer is meditation.

I meditate because I want to meditate. I do not need a salary for it. I do not need to demonstrate that it increases my output.

I practice jiu-jitsu. I enjoy walking. I enjoy comedy. I want to make things that make me laugh, even if nobody else laughs.

I don't need to be the world's best walker to enjoy a walk.

I don't need to beat an AI at jiu-jitsu to enjoy practicing.

I don't need to invent the world's most effective meditation protocol to sit and meditate.

These are not things I do primarily because someone wants to purchase the result.

I do them because I want to do them.

That gives us two questions worth asking:

What would you still do if nobody paid you?

What would you still do if nobody else wanted it?

Not what you would do to become more productive. Not what you would do to gain a skill that makes you harder to replace. Not what you would do to build a defensible personal brand.

What would you do simply because you want to?

You don't have to be the best

For a long time, we have organized our lives around being useful.

Become an expert. Get hired. Build a company. Produce something valuable. Prove your worth.

AGI challenges that arrangement.

There are enormous unresolved questions about who owns AI, who controls resources, and how people will obtain what they need. Intelligence alone does not make food, housing, and energy magically free. Those problems deserve serious work.

But even if we solve them, another question remains.

What is a human life for when we no longer need to justify it through productivity?

You can still do mathematics. You can still write. You can still build. You can still play chess.

You may simply be doing those things as an amateur.

And perhaps that word will no longer sound like a downgrade.

AI will turn every pro into an amateur.

Not the end of activity.

The end of productive activity.

And maybe the beginning of doing things simply because we want to.

---

Editorial note: This is an argument about a possible mature-AGI economy, not a verified prediction that every occupation will disappear on a specific timeline. The twenty-second passage is a thought experiment.

What would you still do if nobody paid you?

What would you still do if nobody else wanted it?