Skills to Learn for the AI Era: How to Prepare for AGI

These are the skills to learn for the AI era, and the knowledge worth building now. The short version: the price of any work AI can do will sink toward the cost of running the AI, and what people earn will shift toward what they own and whether they can be trusted because they’re human. So preparing comes in two stages: skills to get through the transition, and skills to live well after it.

Where the price of labor goes in the AI era

One basic economic rule covers it. If something can do the same job cheaper, the price of the job can’t stay above the cost of that alternative. A translator’s hourly rate can’t stay much above what it costs an AI to translate the same pages, and that cost keeps falling. In the long run it heads toward electricity plus chips.

The best analogy is the horse. In the 1910s the US had well over 20 million horses. Once cars and tractors arrived, what a horse could earn by working fell below what it cost to feed one, and by around 1960 only a few million were left. Horses didn’t get lazy. Their labor became worth less than their upkeep.

The usual counterargument is comparative advantage: even if AI is better at everything, humans still have something relatively better to do. True, but only while AI is scarce. If you can spin up as much AI as you want, comparative advantage doesn’t protect wages. Horses had comparative advantage too.

Some work goes the other way, though.

Work that’s valuable because a human does it: a string quartet needed four players 200 years ago and still does. When everything else gets cheaper, work that can’t get more productive gets relatively more expensive (Baumol’s cost disease).
Hands-on work in the physical world goes last: what’s hard for people (math, analysis) is easy for machines, and what’s easy for people (walking, picking things up) is hard for machines (Moravec’s paradox). Screen work goes first. Plumbing, nursing and hairdressing go much later.

So money moves away from labor and toward things AI can’t multiply at will: capital (chips and models), energy, well-located land, and people’s trust and attention.

We ran this experiment 2,500 years ago

A society that doesn’t need to work isn’t new. Victorious ancient states ran something close to it on slave labor and tribute. That “freedom” was built on other people’s forced labor, so it’s nothing to romanticize, but it shows clearly what a society does with spare time and spare wealth.

Athens used tribute from its allies and silver mined by slaves to pay citizens for jury duty, so even poor citizens could take part in politics. Within a few decades, one city produced democracy, tragedy, philosophy and the writing of history. The Greek word for leisure, scholē, is where “school” comes from.
Sparta had helots, conquered serfs who outnumbered citizens several times over, doing all the farming. Spartans spent their free time on military training and surveillance because they feared a revolt. They left almost no culture behind, and the citizen body shrank from several thousand to under a thousand.
Rome saw small farmers pushed out by slave-run estates and into the city as an urban poor. The Gracchi brothers tried to redistribute land and were killed. Instead, the poor were kept quiet with free grain and gladiator games: bread and circuses. The elite competed through luxury and bribery until civil war ended the Republic. Slaves were so cheap that nobody bothered building machines.

In his Politics, Aristotle wrote something along these lines: if a loom’s shuttle could weave by itself, masters would have no need of slaves. That loom is on its way.

The lesson: what a society does with its spare time decides where it ends up. With the same leisure, Athens participated and built, Sparta feared and controlled, and the Roman poor watched.

Skills for the transition (the next 10–20 years)

1. Defining problems and directing AI

The gap between people who use AI and people replaced by it comes down less to technical skill than to deciding what to build, why, and whether the result is right. I don’t write code, but I built this whole site by telling AI what I wanted. The Apple Game went from a single request to v13 in an afternoon. The AI writes the code, but I’m the one who has to say “no, that’s not what I meant.” That’s the practice.

2. Building ownership, not just labor

Financial assets
Writing, channels and domains that keep working while you sleep
People who trust you and follow you: an audience, a community

The last one is the most underrated. The more content AI can churn out, the scarcer “I believe it because this person said it” becomes.

3. One kind of work that’s valuable because you’re human

The test: if AI could do it as well or better, would people still pay a person?

Work where the relationship is the product: tutors, counselors, coaches, caregivers. This lasts the longest
Work where watching a human is the point: sports, live music, a chef cooking in front of you, handmade goods. Chess engines have beaten humans for almost 30 years, and more people play chess than ever
Work where having actually lived it is the value: reviews from real users, travel writing from someone who went, recipes that include what went wrong
Work that brings people together: club organizers, community managers, the neighborhood shop owner
Work where one person’s taste is the value: curators, DJs, sommeliers
Work where a human doing it is the luxury: fine-dining service, concierges, personal trainers
Work with an expiry date, but a long one: work that’s about bearing responsibility (final sign-offs, rulings) shrinks once laws change, and trades like plumbing, electrical work and nursing last until robots catch up

4. Combining the two

Most of that work doesn’t scale. A tutor can’t teach more than 24 hours a day. So the strongest combination is human-valued work plus a container you own to hold it. Recipes you’ve actually cooked, collected in something like a recipe index. The trust you’ve built tutoring, turned into a word-of-mouth network.

Skills for after the transition

Once work isn’t necessary, temperament matters more than ability.

Generating your own motivation: finding something to throw yourself into when nobody asks and it pays nothing. It’s most of the difference between people who fall apart in retirement and people who flourish
Your body: exercise, cooking, making things with your hands. The more endless digital pleasure gets, the more physical experience anchors you
Bringing people together: when workplaces disappear, so does the biggest place people meet
Taste and judgment: when everything can be generated without limit, what you choose becomes who you are
Being a doer, not a spectator: Athenian leisure wasn’t sitting and watching. It was arguing, competing and getting on stage. In Latin, leisure is otium and business is negotium, literally “not-leisure.” For the Roman elite, leisure was the default

What to study for the AI era

AI already knows every fact you could memorize. But the foundation for judgment has to live in your own head.

Math and logic: to notice when the AI is wrong
Writing: a tool for organizing your own thinking. Telling AI what to do is writing too
History and philosophy: “how do you live without work?” is a question people have worked on for thousands of years. The Stoics, Epicurus and the lives of ancient aristocrats all have hints
How money and the economy work: to hold your ground in the transition, you need to know where money flows

What to avoid

Passive consumption as a habit: a brain trained on endless scrolling is the first thing to crack when work disappears. Rome’s circuses are today’s algorithmic feeds
Hanging your whole identity on one job: if “I’m someone who does X” is defined only by a job, you go down with it

What you can do right now

Start one project you’d do even if it never paid, and take it seriously
Put your name on a record of things you’ve actually done, in writing or on video
Have AI help you build one thing all the way to the end
Build one exercise habit
Join a group, or start one
Read one book of history or philosophy

In one line: during the transition, direct AI and build things that are yours; for what comes after, train the muscle of making your own meaning. All of this is a prediction, of course. But people already showed us this pattern 2,500 years ago, so I don’t think it’s a wild guess.