open allocation infrastructure, for humans and for agents.
The largest pool of liquid altruistic capital in modern history is about to land. Increasingly it will be moved not only by people, but by agents acting on their behalf — recommending, scoring, negotiating, and allocating at a volume the existing institutions were never built to absorb.
Orchestrating that at scale needs an allocation layer that is trustworthy by construction: open, auditable, and reproducible to the dollar — so that many funders, recommenders, applicants, and the agents working for each can coordinate without having to trust a single operator or a black box.
Phil is building that layer as small, composable primitives. The first one is the allocation mechanism itself.
SFF's S-process is a marginal-utility allocation mechanism for grant rounds: funders, recommenders, and applicants each express preferences, and the mechanism settles them into dollar allocations. It works. But the methodology is partly tacit and partly embedded in private tooling.
Open S-process is a clean-room, reimplementable version, built so the mechanism can be shared infrastructure rather than one organization's internal tool:
As more of the grantmaking loop is done by agents, the allocation layer stops being a spreadsheet inside one foundation and becomes a protocol many parties run against. The thing that lets that scale isn't a smarter model — it's a mechanism simple enough to specify, verify, and reproduce, so trust comes from the math rather than from the institution holding it.
Skate there: a credibly neutral, open allocation primitive that humans and agents can both drive, compose, and check.