Paper: whitepaper.md · Audits: whitepaper-framework-audit.md · theory-hardening-audit.md
| ID | Statement |
|---|---|
| C1 | Mutation ∝ compute; convergence ∝ bounded cognition. Propose cheaper than accept. |
| C2 | Delivery follows accepted state, not proposed. |
| C3 | Trustworthy acceptance requires reviewability within policy bounds (sync bandwidth). |
| C4 | Propose-rate > review bound → informational debt; verify without scoped confidence; acceptance without diff grounding. |
| C5 | Local stabilization ≠ global comprehensibility without composition. |
| C6 | Correctness ≠ comprehensibility. |
| C7 | Fixed review coordinates reduce context-reconstruction cost. |
Observation (not invariant): constraint persistence (§1.3)—acceptance disciplines outlast mutation substrates in observed environments.
Boundaries: B1 tests ≠ stabilization · B2 narrative ≠ acceptance · B3 maximize convergence not propose-rate · B4 concurrent epochs accelerate debt.
Strong: propose-rate → bound stress. Medium: bound stress → debt; debt → degradation modes. Chain breaks when accept keeps pace and coordinates stay stable.
Propose overload → stable coordinates · debt → merge authority + gates · opacity → repository truth + review · engine churn → supervision decoupling (§1.3).
Mutation lowers implementation cost faster than coordination cost. Bottleneck may shift to convergence under load (reference observation, §12).
Fragmented multi-operator merge authority weakens C5. Full autonomy out of scope.
One deployment: coordination effort dominant on modest domain—anecdotal. Annex A = embodiment only.
Chat plans. The repo is truth. You merge.