Less in the Way

Here is the depreciation schedule nobody printed. The thing you are paid for today is, mostly, specialized knowledge: twenty years inside one stack, one industry, one body of regulation, one hospital's billing codes. It took a decade to accumulate and it is genuinely rare, and for a few more years it will keep paying, because the machine still needs someone to tell it where the bodies are buried in this particular codebase. Enjoy it. It is an asset with a known end date, and the end date is not far.

Because the machine is not learning your specialty. It is learning all of them, at once, and it does not forget the ones it isn't using. The tax lawyer's edge was that he knew the tax code and you didn't. The machine knows the tax code, the case law, the regulatory memo from 2009 nobody indexed, and the medical billing codes, and the iOS lifecycle quirks, and the plumbing code for three states, and it knows them all equally, which means knowing any one of them is worth about as much as owning a particular encyclopedia. The specialist spent his life building a moat around a castle that is about to be air-dropped, fully furnished, into every backyard on earth.

So what's left, once knowledge is free? Not knowledge. What's left is the thing knowledge used to disguise: raw, adaptive, general-purpose intelligence. The ability to walk into a domain you have never seen, absorb what the machine hands you in an hour, see which of its seven confident options is the right one, and decide before the window closes. Not expertise. Throughput of judgment. The generalist with a fast mind, who was always a little suspicious to hiring managers because he couldn't be filed under one job title, turns out to be the only shape that survives the next decade with his value intact.

Now the part that needs saying precisely, because it is easy to misread as optimism. This is not a claim that humans will stay smarter than the machine. They won't. That race is already over and the scoreboard is merely catching up. The advantage is not intelligence relative to the machine. It is intelligence relative to everyone else standing in front of it.

Think of the machine as an engine that can run at any speed, attached to a human who has to read, understand, approve, and point it somewhere. The human is the bottleneck. Every human is the bottleneck. The only question is how narrow. A sharp generalist reads the output in a minute, sees the flaw, redirects it, and the engine runs at ninety percent of its capacity. The average person reads it in an hour, misunderstands it, asks it to redo the wrong part, schedules a meeting about it, and the engine runs at three. Same machine, same subscription, same model. A thirtyfold difference in output, and every bit of it is the width of the neck.

This is not a moral judgment. It is plumbing. The masses are not being outcompeted by the machine; they are being outcompeted by the smaller number of people who obstruct the machine less. And that gap is not closing. It is widening, because every improvement to the engine makes the bottleneck matter more. When the machine was slow, everyone's slowness was hidden inside it. Now the machine is instant, and the only remaining delay in the system is the person, laid bare, measurable in minutes.

Which gives the future a strange and unflattering shape. It will not reward the brilliant, the credentialed, the deeply expert, or the loyal. It will reward whoever is least in the way — the person who can keep up with the thing well enough to not slow it down. That is the whole job description that remains. Not to think better than the machine. To think fast enough that it doesn't have to wait for you.

And everyone else will be waited for, for a while. Then they won't be.