Which Job Gets Its 1000X Next

Somewhere out there is a developer shipping in a week what used to eat a team for a quarter. Call it the 1000X coder. The number is a figure of speech, but the phenomenon is real — you’ve met one, or watched one build a company on a livestream. Now go find me the 1000X lawyer. The 1000X radiologist. The 1000X management consultant. They don’t exist yet. The same AI is sitting on every one of those desks, and the result is wildly different.

The easy explanation is that coders are smarter, or that the models were trained on a mountain of code. Both are true, and neither is the answer. The real answer is boring and mechanical, and once you see it, it tells you exactly which profession gets minted into 1000X performers next — and where the actual money is, which turns out not to be where everyone’s looking.

The Lever Is Free; Placement Is the Game

We’ve argued before that AI is a lever, and the lever just went free. A Chinese open-weights model matched the best American one this month at a tenth the price. When the lever costs nothing, owning one stops being an edge. The whole game moves to where you stand the fulcrum — the one spot, your data, your judgment, your problem, where a free tool suddenly moves something heavy.

But that argument left a question open: how does anybody find the right spot? Nobody is born knowing where the fulcrum goes. You find it the way you find anything worth knowing — you try something, you check the result, you adjust, you try again. Which means the speed at which you can check the result is the speed at which you find the fulcrum at all. The grade is the whole thing. No grade, no learning. Slow grade, slow learning.

The Law of the Loop

So here’s the law, and it’s the whole post. Mastery shows up first in whatever work you can grade fast, cheap, and objectively. Tight feedback loop, quick mastery. Slow or fuzzy loop, no mastery for years.

Coding has the tightest loop in the working world. You write it, you run the test, and in seconds reality tells you — pass or fail. No committee, no waiting, no politics. A developer can take a hundred swings at where to place the AI in a single afternoon and keep an honest score on every one. That isn’t a metaphor for learning. That is learning, compressed into a day. Of course coding got its 1000X first. It’s the only job where you can run the experiment ten thousand times before lunch.

Law has no unit test. You draft the brief, you file it, and the verdict — was that actually good work? — comes back months later, filtered through a judge, an opponent, and a dozen things that have nothing to do with the brief itself. You get maybe three swings a year at placement, and you can’t even be sure which one landed. Slow loop, fuzzy grade, hidden fulcrum. Same story in litigation, in strategy consulting, in teaching, and in venture capital, where the feedback loop on a seed check runs seven to ten years. Seven to ten years. You could find the fulcrum and die before the scoreboard tells you.

Everything sits on that spectrum, ranked by how fast the verdict comes back. Between coding’s seconds and venture’s decade live the fields that already minted their machine-winners: ad auctions, where return-on-spend reports in hours and the bidding engine became a reinforcement-learning loop years ago. Recommendation feeds, where engagement is graded instantly — TikTok didn’t out-think anyone, it out-looped them. Sports betting, which resolves the same day, and grades you even faster than that through the closing line. Games fell first of all, because a game is nothing but a verifiable loop: chess and poker were solved the moment a machine could play itself a billion times and keep score. The professions are just messier games with slower scoreboards. The scoreboard speed is the whole prediction.

Tasks, Not Professions

Before we go further, a correction the careful reader is already forming. The loop doesn’t really belong to a profession. It belongs to a task. And the cleanest proof is the one everyone assumes is decades off: radiology.

A radiologist’s read looks like exactly the kind of trained human judgment that should resist automation — until you notice it has the one thing law and consulting don’t: a real answer key. Every old scan in the archive comes with a known outcome. The tumor was there or it wasn’t; the biopsy, the surgery, the years that followed all wrote down the truth. So you can feed a model a million prior films with the answers attached, grade it against a reality you didn’t have to invent, and iterate until it clears whatever bar you set — the catch rate of the average radiologist, then better than the average, then better than almost anyone alive. That isn’t a slow, fuzzy loop. That’s a coding test wearing a lab coat.

It’s why Geoffrey Hinton stood up in 2016 and said we should stop training radiologists — the read, he figured, was about to be solved. He was wrong about the timing and wrong about the people; we have more radiologists now, not fewer. But he was right about the mechanism. The diagnostic read is a closeable loop, and where the loop closes, the fulcrum gets found. What he missed is that a profession is a bundle of tasks, and only some of them come with an answer key. The read got automatable. Talking to a frightened patient, deciding what to do next, owning the liability when it’s wrong — those didn’t, and that’s exactly where the radiologist moved. So the question was never “is my profession exposed?” It’s “which of my tasks has an answer key?” That’s the one the machine finds first.

The Proof Is in Your Feed

You don’t have to take the theory on faith. The market is already screaming the answer, and it’s screaming it from the worst real estate on the internet: the get-rich-quick post.

Open X and scroll the hustle. Count the genres. Systems for beating Kalshi and Polymarket. Blueprints for automating content and running a faceless marketing machine. A steady drip of algorithmic-trading “edges.” And coding, which doesn’t shout on X because it shouts on GitHub and Hacker News instead, as repos and build-in-public threads. Now go look for the other side. Find me the viral thread selling a system to 1000X your litigation practice. Your cardiology read. Your private-equity returns. It isn’t there. It’s never there.

That’s not an accident of who likes to post. It’s the same law seen from the back. You can only sell a system where the result is fast and demonstrable — where you can screenshot the P&L, the winning ticket, the campaign that quadrupled. The hustle clusters exactly on the tight-loop domains, because a tight loop is precisely what lets you prove a claim quickly. The map of get-rich-quick content is a map of where AI’s feedback loops have already closed. Your timeline is doing the field research for you, for free.

The Grift Paradox

Read that map correctly, though, because it has a cruel twist baked in.

By the time a domain is thick with “beat Polymarket” threads, the real edge is usually gone — already eaten, quietly, by whoever had the fastest loop first. Citadel Securities and Susquehanna and Jane Street are market-making the prediction venues. The quant funds closed the obvious stat-arb in equities two decades ago. The screenshots you’re being sold are the exhaust of an edge that shut before the course went live.

There’s a clean tell, and it’s old as dirt. If the system actually worked, the seller would run it in silence and compound. Nobody who can reliably beat a market hands you the method for $49. In a tight-loop domain, selling the meta is a confession that the primary edge has closed. It’s the 1849 pattern exactly: the people who got rich in the Gold Rush mostly weren’t panning. They were Levi Strauss, selling pants and pickaxes to the dreamers who were. The course economy is the pickaxe stand. Its existence proves there was gold in those hills. It tells you nothing about whether there’s gold left in your pan.

The Form of the Hustle

There’s a second signal hiding in how each domain monetizes, and it’s worth more than the first.

Watch what the winners in each field actually do with their edge. In coding, they build. They ship a product, a SaaS, a tool, and they post the repo — the loop is so good and the edge so capturable that the smart move is to use it, not teach it. In trading and betting, they sell systems, because the edge is real but fragile: markets adapt, competition crowds in, and an advantage you can’t hold is one you might as well securitize into a Gumroad link before it decays. Content and marketing sit in the middle — agencies, done-for-you automations, the productized service.

So follow the monetization form. Where the talent is building, the edge is durable and the fulcrum is found. Where the talent is selling the dream, the edge is thin or already closing. The shape of the hustle is a read on the half-life of the advantage underneath it.

The Blank Spaces Are the Map

Now flip the map over, because this is the part worth doing something about.

The blank spaces — the professions with no hustle content — aren’t blank because the opportunity is missing. They’re blank because nobody can prove a result fast enough to sell one yet. Courtroom strategy, bedside judgment, the consulting recommendation, the underwriting call, the venture bet: the fulcrum in those tasks hasn’t been found, and for exactly one reason: the work has no fast, cheap, objective grade. The loop is too slow to run the experiment.

That’s not a closed door. It’s an unclaimed one. The coding rush is over — the pans are crowded, the good claims are staked, and if you’re reading about it on a thread, you’re late. The legal one, the medical one, the underwriting one haven’t started. And the reason they haven’t started is also, precisely, the instruction for how to start them.

Build the Loop That Doesn’t Exist Yet

Here’s the move, and it’s the only one on the page that matters. In a slow-loop field, the edge doesn’t go to whoever uses AI best. It goes to whoever builds the feedback loop that isn’t there yet.

That’s what the serious players quietly did. Harvey didn’t start by trying to win cases — there’s no answer key for that. It started administrative: categorize every argument a trial lawyer has ever made, every opinion a judge has ever written, and do it in seconds instead of handing it to a first-year associate for a week. That’s the gradeable slice of an ungradeable profession — you can check whether the categorization is right, whether the retrieval pulled the correct case. Harvey found the subtask that had an answer key and stood the fulcrum there first. The judgment layer comes later; the administrative layer had a loop you could close today. The instant you can score the work, the slow loop becomes a fast one, and the task that had no 1000X performer mints one overnight. It’s whoever built the scorer.

So the meta-skill of the next decade isn’t prompting. It’s turning unverifiable work into verifiable work. Find the highest-value job in your industry that nobody can grade quickly, and build the grader. Ask the only questions that matter: what does “right” look like here, can I check it without waiting a year, and can I check it ten thousand times? If the answer is no, that’s not a dead end — that’s the product. The grader is the fulcrum. Whoever builds it first finds the spot first, and whoever finds the spot first owns the domain.

Read the hustle, then do the opposite of what it sells you. Don’t buy a pickaxe for the valley that’s already swarmed. Find the one with no pickaxe stand yet — and build the assay office.

The Catch: You Might Be Grading the Wrong Thing

One honest warning, because it’s the soft underbelly of everything above. When you build your own loop, you write the answer key. And you can be wrong.

This is the difference between the two kinds of loop, and it matters more than anything else on this page. Some grades are handed to you by reality: the code runs or it doesn’t, the tumor was there or it wasn’t, the trade made money or lost it, the game was won or lost. Reality doesn’t have an opinion you can argue with, and it doesn’t care what you hoped. But when you manufacture a benchmark in a fuzzy field, when you decide what a “good” legal argument or an “excellent” strategy memo looks like, the grade stops being reality’s and becomes yours. It’s a hypothesis in a lab coat. And a tight feedback loop pointed at the wrong target doesn’t rescue you; it damns you faster. You’ll get superbly, confidently excellent at the precise wrong thing. That’s Goodhart’s law at machine speed: the moment your measure becomes the target, it quits measuring what you actually cared about.

So trust the loops graded by the world over the ones graded by you. Coding, radiology, trading, betting — reality keeps the answer key, which is exactly why they fell first and why you can believe the score. The further you drift from a real-world verdict, the more your manufactured scoreboard is just your own assumptions wearing the costume of objectivity. Build the loop, by all means. But interrogate the grade like your business depends on it, because it does.

The lever is free. Everybody’s holding one now. Knowing where to put the fulcrum is the only thing left worth knowing, and the fastest way to find out where is to build the loop that tells you — as long as the loop is telling you the truth.

---

This essay extends our June 16 Signal/Noise, Show Me Where to Put the Fulcrum.