zas

Market infrastructure for outcome finance.

Funding models were built for a world that moved slower than this one. They decide once, at the start, and then go quiet until the result arrives — if it arrives. We think it can be continuous instead: valued as it goes, by everyone with something to contribute, and corrected as results come in. Science first, for reasons below.

Financing what actually matters

Capital reaches the places it already knows. Science funding was designed in 1951, when most research money was public; the ratio has since inverted, and the apparatus for allocating it has not moved. Committee review is slow, gatekept, and silent between the application and the result. That is a coordination problem before it is a finance problem.

We build market protocols for the parts nobody wants to own: measurement, verification, settlement, and the negotiation of who owes what to whom.

Measurement is the hard part

Every claim about an outcome is a claim about the world, and the world is expensive to observe. A proof establishes that a claim follows from the data by the stated methodology. It does not establish that the data reflects the field, and anyone who tells you otherwise is selling something.

What it does establish is worth more than it sounds: a claim can be proved against a threshold without publishing what sits underneath it, and bound to a version of a methodology, so quietly revising any encapsulated workflow breaks the proof. Someone stands behind it, with a limit, and can be pursued if it fails. Once a claim is bound and settled against a real outcome, you have something to learn from.

“It’s not just human intelligence, it’s as much about the collective as it is about the individual. … We could be creating new kinds of collectives that do new things — not the super intelligence replacing a human.”

— Michael I. Jordan, An Alternative View on AI: Collaborative Learning, Incentives, and Social Welfare, Stanford Data Science Distinguished Lecture Series, 2023

Intelligent market mechanisms, not products

Intelligent in Jordan’s sense, not the other one. Nobody here is building a mind. The intelligence is in the arrangement — what people and machines can work out together that none of them could alone — and the mechanism’s only job is to make that arrangement honest.

It also learns. Every market that resolves tells us something about how well we valued it, and that correction goes back into the mechanism rather than into a report nobody reads. It gets less wrong as evidence arrives. That only works where outcomes actually resolve, which is a constraint on what we build and where we start.

A mechanism still earns its keep by being uninteresting. Legible enough that a regulator can read it, cheap enough that a small cooperative can run it, dull enough that nobody needs to trust us specifically. Open specifications, reference implementations anyone can fork, and economic rules that hold when the participants dislike each other.

Science first

Because science resolves. Not all of it, and not quickly — but a replication works or it does not, and you find out in months rather than decades. That is the condition a mechanism needs in order to learn how well it values things. It is also, at present, the one field where institutions will pay to improve how the funding itself works.

Turnover funds the work directly. More trading means more experiments, and more, smaller, faster experiments is what produces knowledge. That alignment is structural; we did not have to argue ourselves into it.

Where this goes

Smart funds that disburse against verified outcomes. Registries that federate instead of consolidating. Disputes settled by people drawn at random from those holding a position, advised rather than instructed by an expert. None of it novel on its own; the work is in making the pieces interoperate.

If any of that sounds like your problem too, the door is open.