My agents can now do almost everything I hired myself to do. What is left is stranger and harder to name.
I went looking, the other morning, for a task that was still mine. I needed a competitor teardown — pricing, positioning, the gaps in their onboarding, a read on where they were vulnerable. Two years ago that was a good day's work and I'd have felt useful doing it. Now I described what I wanted in four sentences, walked the dog, and came back to a document better than the one I would have written, with sources I wouldn't have thought to check. The same thing happened with the financial model. And the hiring rubric. And the first draft of a contract that my lawyer red-lined in two places and otherwise left alone.
I am, by training and temperament, an analyst. I built my early companies on the back of spreadsheets I was quietly proud of. And the machines have come for the spreadsheets. They have come for the memos, the research, the synthesis, the careful weighing of options that I used to think was the job. They do it faster, and on a good day they do it better, and they do it without the small resentments that used to leak into my Thursday afternoons.
So I had to ask the question that everyone in this business is circling and few will say plainly: if the analysis is done, what am I for?
The engines got cheap
Here is the shape of what happened, because the shape matters.
For most of the history of work, the bottleneck was the engine — the capacity to do the thing. To produce the analysis, write the code, draft the brief, run the numbers. Engines were scarce and expensive, so we organized everything around them. We hired engines. We trained for years to become better engines. The whole apparatus of the modern firm — the org chart, the headcount, the review cycle — is a machine for renting and coordinating cognitive engines.
In 2026 the engine got cheap. Not cheap as in discounted. Cheap as in abundant, the way air is abundant. A founder I know runs what is, in any honest accounting, a forty-person operation with four people and a swarm of agents. The agents draft, research, model, test, and revise. They don't tire and they don't sulk. The output that used to require a department now requires a prompt and a little patience.
When something that was scarce becomes abundant, its value doesn't fall gently. It collapses. And the thing that was abundant — that everyone overlooked because it was lying around for free — becomes the whole game.
Think of a ship in the age when wind was the only engine. Everything turned on the sails: the size of them, the men who could work them, the luck of the weather. A captain's craft was mostly the management of that scarce, temperamental power. Now imagine engines become free and limitless. Every ship can go anywhere, fast, in any wind. You'd think this makes the captain more powerful. It does the opposite to most of them. When propulsion is free, knowing how to propel is worth nothing. The only thing left that's worth anything is the rudder — the decision about which shore to sail toward, and why, and whether that shore was worth the crossing at all.
The engines got cheap. Something still has to decide which shore.
The thing no model can hold
You might say: fine, but the machine can choose the shore too. Ask it which market to enter, which feature to cut, whether to take the money on the table, and it will give you an answer — reasoned, balanced, often shrewd.
It will. And you should listen. But notice what it can't do, and why.
A model can tell you the consequences of sailing to a given shore. It cannot tell you whether that shore is worth your one finite life. It has no life. It will weigh your options with perfect equanimity precisely because it has nothing riding on the outcome — no children, no mortgage, no decades of wondering what you were actually built for, no particular grief it is trying to outrun, no shore it has been quietly dreaming of since it was nineteen. It is a magnificent advisor and a hollow principal. It can hold the map. It cannot want the destination.
This is the part that took me a while to see clearly, because it sounds soft and it is in fact the hardest thing here. Judgment about means — what's the most efficient route, the cheapest, the fastest — is analysis, and analysis is now a commodity you can buy by the token. Judgment about ends — which problem is worth solving, which life is worth building, which thing the world will be better for having — is not analysis. It's taste applied to stakes. And taste, the real kind, only forms in a creature that can lose.
A man at a poker table plays differently with his own money. Everyone knows this. We don't usually notice that it's the deepest fact about him as a decision-maker, not a flaw in his composure. The stakes are what make the judgment real. Strip them out and you don't get a purer, cooler version of the same judgment. You get advice. Good advice, sometimes. But advice is what you collect before the decision; it is not the decision.
What this asks of us
I have a restless mind. For most of my life I treated it as a defect — the inability to stay on a path, the way every project I started would lure me sideways toward a more interesting one before I'd finished the first. It took me an embarrassingly long time to understand that the restlessness wasn't a bug to be managed. It was a direction-finder. The part of me that couldn't stop asking but is this the right thing to be doing at all was, it turns out, the one part of the job that wasn't going to be automated.
That reframe is no longer just personal. It's the strategic situation of every knowledge worker alive.
For two centuries we have trained people to be excellent engines. The schools optimize for it, the firms reward it, the whole status ladder of professional life is a ranking of who is the more reliable analyst. We are about to discover that we trained for the wrong scarcity. The engine is free now. The premium has moved — quietly, and faster than most institutions will admit — to the faculty we never bothered to teach because we assumed everyone had it lying around: the ability to decide what is worth doing, to bear the weight of that decision, and to have the taste to know a good shore from a merely reachable one.
This is not a comforting conclusion and I won't dress it up as one. A great many people built careers on being the best engine in the room, and the room no longer needs the best engine. But there's a harder and better thing on the other side of that loss. When the analysis is free, you are finally allowed to spend your scarce attention on the question that actually deserved it the whole time — not how do I do this, but should this be done, by me, now, at the cost of everything else I might have done instead.
That question was always the human one. We just hid it, for two hundred years, behind the busywork of being engines. The machines have taken the busywork. What they've left us is the rudder, and the strange terror and privilege of having to steer.
Decide where the ship is going. That part is still yours. It may be the last part that ever was.