I want to give the people whose work this stands on their due, and to be plain about what is mine.
The base idea of several knowledge sources reading and writing one shared structure belongs to the blackboard architecture of the middle 1970s and to the tuple-space coordination that grew up beside it, and both are good ideas I did not have. More recently, an LLM-based multi-agent blackboard system posts requests to a shared board that autonomous agents volunteer to answer by capability, without a central coordinator. That teaches the volunteer-coordination pattern and only that: it has no per-record governance, no sovereign participant-held context, no recursion to mastery, and fuses none of them. A goal-oriented tutoring framework maps a learner's goal to skills, finds the gaps, and schedules a path. That teaches the learning loop and only that: it composes no sovereign context held out of institutional custody and governs nothing at the data layer. Capability-scoped remote interfaces are known in general; the point here is their use as the means by which heterogeneous seats coordinate on governed shared state without host access. User-held data stores with access control at the data layer exist too, Solid and its pods being the best known, and they are the right idea; what none of them does is compose a held context with an authority's goal under a loop that persists to mastery, across many sovereign seats, under one rule.
Will Ruddick and Grassroots Economics have spent more than a decade in Kenya building commitment pooling, and I am going to let them describe it, because it is theirs to describe: commitment pooling lets communities create, manage, and connect their own economic systems. I met Will at a conference and he is doing the real thing. I will not characterize his work further than he does; go read them. My market is a different case. Here in the United States, scarcity is not the problem. The materials are everywhere. Food rots on loading docks three miles from people who need it, and what is missing is not a medium but coordination, and coordination does not need a wallet. The food pantry that wants to know who gave what, the neighbor who wants to remember who fixed the fence, the donor who wants a receipt, none of them want a ledger. They want a thin record. So this method records non-circulating cooperation, who gave and who received, which never changes hands again and needs no voucher, no settlement, and no balance to reconcile. That removes the reason for the apparatus and leaves a relational memory in its place. Will's work and mine are answers to different questions in different places, and where what a community needs is what Will builds, his work is where I would send them. Record-level access control is old and standard, used here as the coordination-governance boundary rather than as tenant isolation, and as one leg of a combination rather than alone.
None of these, alone or together, teaches the interlock described here. I take no credit for the blackboard, the tuple space, the learning loop, the learner profile, the voucher, or the scoped interface. I built the running system before I knew the name of the blackboard architecture, and I name it, and the nearer recent work above, so that credit lands where it belongs. What is mine, and what this document gives away, is the single rule that makes the three legs one thing, and the coordination that rides on it. It is published so that it stays available to everyone, and so that no one can later fence it off.