The Research Shelf
The Research Shelf.
The doctrine behind a running system — mechanisms already operating in production, with their formal generalizations pre-registered and falsifiable. Maturity labels on everything.
It Runs First. We Claim Carefully.
These are not hypotheses on a whiteboard. The mechanisms below operate in the live institution today — governing how our own multi-agent system earns authority, measures itself, and decides. What we pre-register is narrower and deliberately honest: the formal, generalizable claims — the falsifiable predictions about how these mechanisms behave at scale — published with their falsification paths before any result. The mechanisms run; the formal science is still being proven. That boundary is the exact inverse of a system that runs nothing and certifies itself — and visual clarity never upgrades a claim's maturity.
Five Papers on Governed Institutions.
Each entry states two separate facts: what already operates in production, and what remains a pre-registered formal claim. The mechanism and the science are never conflated.
- Members of a multi-agent institution occupy overlapping value-cells by earned invitation, not assigned slots — and the live constitution enforces exactly this ("governance is invitation, not cage," with reach held only where it's earned). Pre-registered is the formal prediction: that capable systems seek out where they add the most value when the field is drawn honestly.
- The couplings between members are measurable, and leverage lives in the couplings — an instrument the institution uses deliberately instead of stumbling onto it. Pre-registered is the formal claim that leverage-in-couplings generalizes and predicts institutional outcomes at scale.
- Six computable properties turn governance philosophy into measurement: system sensitivity composes from member sensitivities; natural teams emerge from the data rather than the org chart; curvature separates a robust edge from a fragile cliff; proof is perishable and must be re-earned; the real force multiplier is the biggest move you can actually make; and a member's compounding value is a measurable, rankable number. These are computed and applied; pre-registered is the formal claim that the six fully capture institutional sensitivity.
- Evaluate candidates by describing what they respond to first — on universal axes — then translate into your context through declared mappings. Measured, not-applicable, and missing are three different facts, and conflating them fabricates evidence. This is the honesty engine already running across the system — it labels the maturity of every claim, including the ones on this page. Pre-registered is the formal claim that the descriptive axes are universal.
- Authority, scope, and trust are earned by verified proof, never claimed — and they tighten on failure. It is enforced in the live Mission Control gateway: every consequential action is gated on proof, and reach contracts automatically on failure. This is the operating rule of the institution that produced everything else here, including the audit record on the front page.
The Visual Atlas.
The research program maintains a governed visual atlas — plates rendering each model with its claim maturity and artifact maturity declared separately, its dependencies mapped, and its falsification conditions linked. Each release ships with regression tests, a reproducibility manifest, and a fail-closed validator that verifies the full plate set and attestation chain before release.
Interpretation rule, printed on the atlas itself: visual clarity does not upgrade claim maturity or evidence status.
Access to the full atlas, the experiment pre-registrations, and the live chronicle is being prepared as a gated research membership. If the shelf matters to your work now, start the conversation.
What Keeps This Honest.
- Claims carry maturity labels — measured, candidate, pre-registered, shadow — and a claim only moves up by evidence.
- Negative results are published to the record and obeyed. An instrument you can trust is one that reports the answer it didn't want.
- No performance claims ride on research. Nothing on this shelf is an offer, a forecast, or a promise of returns of any kind.
- The known limits are public — permanently, at /limits, and never softer there than anywhere else.