The Grown-Ups We Hire After the Fact

Ben Bernanke steered the Federal Reserve through the 2008 crisis. This week he joined the independent oversight trust that governs Anthropic, one of the handful of companies training the models that a growing share of the economy now runs on. A man who spent his career deciding how much trust to put into a system that nobody fully controls has been handed the same job again, in a different building, with a different kind of money at stake.

It reads at first like a credentialing move. A famous name on a masthead, the reputational equivalent of a good suit. But the more interesting thing is what the hire admits. You do not recruit a former central banker to babysit a chatbot. You recruit one when you have quietly concluded that the thing you built behaves less like a product and more like an institution — something with systemic weight, something whose failures spill outward onto people who never agreed to use it. The oversight trust is a confession dressed as a governance upgrade. We made something big. We are now looking for the grown-ups.

That pattern is everywhere in this stretch of the industry, if you tilt your head. The same week, JPMorgan signed a deal to become a marquee partner of the NBA. A bank stitching itself deeper into culture, an AI lab stitching itself into the machinery of oversight. Both are institutions reaching for the legitimacy the other one already has. Finance wants to be loved. Technology wants to be trusted. Each is buying a little of what it lacks, and neither can manufacture it from the inside.

The craft outgrows itself

There is a smaller signal underneath the headlines that points the same direction. People who build with these models have started asking whether prompt engineering — the fussy art of phrasing a request just so — is a skill you outgrow. A year ago it felt like the whole game. Say the magic words in the magic order and the machine performs. Now it reads like early scaffolding, the training wheels of a field that is learning to ride.

The honest answer is that you do not outgrow prompt engineering so much as you promote it. The clever phrasing gets absorbed into systems — into evaluation harnesses, guardrails, retrieval pipelines, the boring plumbing that turns a party trick into infrastructure. What looked like a craft was actually a symptom. It was what you did before the discipline existed. The moment a practice matures, its early cleverness stops being a differentiator and becomes a baseline everybody assumes. That is not a loss. That is the sound of a field growing up, and growing up always feels a little like being demoted.

Hold those two things together and a shape appears. The technology is maturing faster than the language we use to govern it. We reached for prompt tricks before we had systems. We are reaching for oversight trusts before we have law. In both cases the improvisation comes first and the structure arrives late, out of breath, hiring former central bankers and building evaluation suites to catch up with something that already shipped.

The children you did not plan for

There is an old story worth keeping in mind here, and it is not a flattering one. The trickster god of the northern myths fathered children with more partners than he could keep track of, and the children were not small. One was a wolf so large the other gods had to chain him. One was a serpent that circled the whole world. The trouble was never the act of creation. The trouble was that the offspring grew, and kept growing, and eventually could not be governed by the hand that made them. The makers spent the rest of the story building restraints for things they had cheerfully brought into being.

That is roughly where the AI industry sits. The creation was the easy part, the exhilarating part, the part everyone got funded for. The governance is the wolf you did not plan for, larger every quarter, and the chains are being forged in a hurry by people who are honest enough to know they are behind. An oversight trust with a former Fed chair on it is a chain. A good one, maybe. But you only forge chains for a thing you already suspect is stronger than you.

None of this is cause for alarm, and I want to be careful not to slide into it. Hiring the grown-ups is the responsible move. It is genuinely better to invite a Bernanke in early than to wait for the serpent to circle the world. The people building these systems are, on the whole, taking the systemic weight seriously, and that seriousness is real and worth naming plainly. The discomfort is not that they are failing. It is that the order of operations is fixed. We build the powerful thing, and then we go looking for someone wise enough to watch it.

There is an old line that people work for three reasons: to survive, to save, to serve. Institutions run on the same three. Anthropic and JPMorgan cleared survive and save long ago. What this week was really about, underneath the press releases, is the third one — the scramble to look like you can be trusted to serve, before the world decides for itself whether you can. The grown-ups we hire after the fact are not there to make the thing safe. They are there to make us feel it might be. Sometimes that feeling arrives before the safety does. Sometimes it arrives instead of it.