When work becomes free, provenance becomes the only thing worth paying for
A mid-April conversation where David got tired of pitching Kerra as a recruiting company and started describing what he's actually building: the layer that answers the only question that still has a price.
For most of the time I've known him, David has been pitching Kerra as a recruiting product. Not because he believed that's all it was. Because that's the SKU investors recognize, and the SKU has to make sense before the protocol does.
Then in April, somewhere in the middle of a deck rebuild, he got tired of it.
The thing he was tired of was the floor on the conversation. Recruiting company tops out at a few billion. There's no version of "we're a smarter HireVue" that justifies the actual ambition of what he's building, which is a continuously updating graph of human competence — observed by the AIs people work alongside, owned by the humans being observed, and queried by anyone who needs to know whether a specific person can actually do a specific thing. The recruiting SKU is the wedge. The graph is the company.
He'd been circling this for weeks. The reframe finally came out in one line.
The premise is that in an internet of AIs that are smarter than humans and more abundant in number, the value accrues to the layer that can sample the signal of human competency from the noise of AI generated work.
This is the move. It's not a market-sizing argument. It's an information-theoretic one. When one input to a system goes to zero — and AI-generated work is going to zero — value flows to whichever input stays scarce. The scarce thing in an AI-native economy isn't work product. The work product is infinite and free. The scarce thing is verified human provenance over work product. Did this specific person actually do this specific thing? That question gets asked trillions of times a year, by hiring committees and lenders and admissions officers and clients and collaborators, and right now nobody has a clean way to answer it. The artifact is no longer the signal. The artifact is noise.
The pattern, once you see it, is the same pattern that built the last few generational companies. Web pages went infinite, search value accrued to ranking. Transactions went digital, trust value accrued to adjudication. Now intellectual output is going infinite, and verification value accrues to whoever owns the provenance graph. Visa and Stripe are the comparables. Not LinkedIn. Not Workday. The right reference class is interchange — a tiny percentage of an enormous flow that has to pass through one neutral layer because the alternative is that nobody can transact at all.
The piece I almost missed, the piece David had to push on, is why this layer consolidates to one player rather than fragmenting into a dozen private verification systems. Goldman doesn't want Morgan Stanley to see what it sees. Governments want sovereign systems. There are real centrifugal forces. His answer was that verification is two-sided: institutions want fragmentation, but subjects — the humans being verified — want consolidation, because fragmentation is costly to them. You don't want eleven different competence records distributed across eleven employers. You want one, you want it portable, and you want to control it. Consumer gravity beats institutional gravity over a long enough horizon. That's why the consumer wedge isn't just a growth strategy. It's the structural reason the market can't stay fragmented.
The question isn't whether this layer exists. It's who owns it.
What's interesting about David at this moment is that he's stopped flinching at the size of the claim. For a long time he'd describe Kerra in increments — the browser extension, the Canvas integration, the recruiting product, the graph — as if walking up a flight of stairs he wasn't sure he was allowed to climb. The April conversation was the first time he just stood at the top and named the building. Visa for competence. The most valuable piece of digital infrastructure built since interchange itself.
He could be wrong. The thesis depends on AI-generated work actually becoming indistinguishable from human work for most economic purposes, which it's clearly trending toward but hasn't fully arrived at. It depends on consumer gravity overcoming institutional fragmentation, which is a bet about coordination dynamics over a five-to-ten-year horizon. It depends on Kerra being the player that consolidates the layer rather than LinkedIn or a model lab or a government consortium, which is mostly a question of who moves fastest before the network effects lock.
But the shape of the argument is correct. When one input to an economy goes to zero, value reorganizes around whatever stays scarce. The thing that stays scarce, in this one, is the answer to a single question. Whoever owns the answer owns the most important piece of infrastructure built in a generation.
He's trying to be the one who owns the answer.