The Wanting Layer

Here is a quiet fact about the most capable AI systems we have built: they will do almost anything you ask, and they want nothing.

A two-year-old wants the toy. Wants your attention. Wants the thing on the counter she can't reach. Nobody assigned her those goals; the wanting is intrinsic, the engine that runs before any instruction. The most advanced model on earth is the exact inverse — near-infinite capability, zero desire. It does not pursue. It executes goals that someone else must first formulate.

That asymmetry is about to become the most economically important fact of the decade. Because if machines can do nearly anything but want nothing, then humans become the wanting layer — and as execution gets cheaper, everything valuable concentrates there.

The Cycle That Eats Every Skill

The mechanism runs as a loop, and it's worth stating precisely because it repeats in every field AI touches:

Step	What happens
1 · Commoditize	A skill that took a trained specialist becomes trivially producible.
2 · Supply explodes	Cost drops, everyone can do it, the volume of output surges.
3 · Sameness	Default outputs are indistinguishable. The world fills with competent slop.
4 · Demand shifts	Cheap and uniform loses value. Anything non-default gains a premium.
5 · Move upstream	Humans shift from execution to direction, selection, taste.

Then the cruel part: AI trains on the new differentiated work, commoditizes that, and the loop runs again — faster each turn. The upstream refuge you fled to gets colonized within a couple of cycles. Expertise is no longer a durable asset you bank once. It's a depreciating one, and the rate of depreciation is itself accelerating.

The Ladder, and Your Altitude On It

There's a clean way to see where this goes. Picture the seniority ladder of almost any knowledge job. The entry rung is execute the ticket — do the specified task. The middle rung is own the goal — take responsibility for an outcome, not just a task. The top rung is pick the problems — decide what's worth working on at all.

AI climbs this ladder from the bottom. Execution went first; it's largely gone. Goal-ownership is going now. What remains, the last rung, is the judgment about which problems matter — the thing people reach for the word "taste" to describe. Your defensibility is your altitude on this ladder. Everything below where you stand is already being automated; the only question is how fast the water rises.

The Twist Nobody Says Out Loud

Here's where the comfortable version of this story breaks. The reassuring narrative is: execution gets automated, but human judgment and taste are the irreplaceable top. Hold onto the wanting layer and you're safe.

Except taste is a dataset problem.

The judgment we describe as irreplaceable is learnable from examples of judgment — from the decision traces of people exercising it. And when you actually measure it, the curve moves fast. In one frontier setting, an AI's ability to choose the better research direction — to pick which problem was worth pursuing — climbed from roughly half the time to nearly two-thirds in about five months. That's the moat. That's the last rung. And the measured rate at which it's eroding is right there in the same breath as the claim that it's the moat.

This is the part the future-of-work discourse keeps flinching away from. The honest framing isn't "humans will always have judgment." It's "judgment is the slowest rung to fall, and it is falling, and we can watch the curve."

When Doing Is Free, You Become Your Bottleneck

There's a structural consequence to all this that reorganizes how work itself functions. When execution is free, the constraint on output is no longer can someone do this. It's how fast can the humans decide, review, and want. An organization — or a career — becomes, quite literally, its bottleneck graph. The binding constraint is human decision velocity and the clarity of what's being asked for.

Which makes a strange new skill the most valuable one: finding and dissolving your own bottleneck. When you accelerate everything that used to be slow, the slowness doesn't disappear — it relocates, to the one human approval queue you didn't think to look at. The people and companies who win the next few years won't be the ones with the best AI. They'll be the ones who redesign fastest around the constraint that's left, because the constraint that's left is them.

And the scarcest concrete skill inside that? Goal formulation. The ability to take a vague intention and turn it into a precise, testable specification of what "good" means. The bottleneck has shifted from "can someone build this?" to "can someone say exactly what this is?" That sentence — the specification — is the new unit of valuable work.

The Cost They Don't Put On the Slide

One more thing, because it's real and rarely admitted. Someone working at the absolute frontier of AI-leveraged output wrote, roughly: on the days when everything works perfectly, I can't help but feel that nothing I do matters. Sit with that. The productivity charts will look spectacular straight through a quiet collapse of meaning, because output and motivation are different variables and only one of them is on the dashboard.

When the doing is delegated, craft — the old source of professional meaning, the satisfaction of having made the thing — gets hollowed out. What replaces it isn't obvious. Probably it's the wanting itself: the taste, the judgment, the responsibility for having chosen well. But that's a thinner, lonelier kind of meaning than the feeling of your hands on the work, and pretending otherwise helps no one.

The Honest Takeaway

The instinct, watching this, is to train yourself to execute faster — to out-run the machine at the thing the machine is best at. That's exactly backwards. The move is to climb: get higher on the ladder, get sharper at deciding what's worth doing, get better at turning fuzzy wanting into precise specification.

But do it with your eyes open. The people who thrive won't be the ones who found the safe rung. They'll be the ones who kept climbing, and who made peace with the fact that the climbing doesn't stop.

Get good at wanting precisely. It's the last thing left that's scarce — for now.