escrow: Aurora autonomous flywheel thesis — defined home + expiration + one-test gate + no-multi-reviewer rule (cross-AI converged refinement, Aaron 2026-04-29)

docs/backlog/P3/B-0094-escrow-aurora-immune-governance-flywheel-thesis-until-prototype-passes-aaron-2026-04-29.md

@@ -0,0 +1,219 @@
+---
+id: B-0094
+priority: P3
+status: escrowed
+title: Escrow Aurora Immune Governance flywheel thesis until prototype passes
+tier: research-deferred
+effort: M
+ask: Aaron 2026-04-29 via Amara translation — *"This is not rejected. It is escrowed. The thesis is important enough that we do not let it land sloppily."*
+created: 2026-04-29
+last_updated: 2026-04-29
+composes_with:
+  - B-0089
+  - B-0093
+  - PR-707
+tags: [aaron-2026-04-29, escrow, research-deferred, aurora-immune-governance, flywheel-thesis, mimetic-coupling, falsifier-gated, session-closure-rule, restraint-discipline]
+---
+
+# B-0094 — Escrow Aurora Immune Governance flywheel thesis until prototype passes
+
+## Source
+
+Aaron 2026-04-29, via Amara's translation of Aaron's intent
+after the multi-AI Aurora-Immune-Governance synthesis arc:
+
+> *"This is not rejected. It is escrowed. The thesis is
+> important enough that we do not let it land sloppily.
+> Don't bury it, don't bloat it, and don't rush it. Put it
+> in escrow and make it earn its way into the system."*
+
+The harsh pushback from external-AI second-opinion-reviewer
+across two consecutive rounds was protecting the thesis,
+not killing it: the synthesis was diagnosing the exact
+pathology it was exhibiting (`r_flywheel_capture` —
+momentum increasing while evidence quality flat). The
+disciplined response is to preserve the thesis without
+integrating it before the prototype runs.
+
+## Why P3 + escrowed status
+
+Escrow is the bounded-preservation status, not the
+deferral-of-work status. The thesis components survive
+session boundaries via this row + the conversation-log
+pointers below. The thesis cannot be lost; it also cannot
+land prematurely.
+
+## The escrowed thesis (verbatim)
+
+```text
+The factory is a network of autonomous desire-bearing flywheels.
+
+Girard / Wanting explain how desire, imitation, rivalry,
+scapegoating, and shared common-good pursuit move through
+the network.
+
+Infer.NET-style probabilistic inference gives the
+computational shape for modeling correlated belief, mimetic
+coupling, and evidence updates.
+
+The immune membrane supplies the action layer: detect
+danger, gate execution, quarantine, require proof, record
+memory, and prevent thin/rivalrous flywheels from capturing
+the factory.
+```
+
+Short form:
+
+```text
+The first sub-thesis names why consensus can lie.
+The second names what the desire is aimed at.
+The third names how to model correlated belief.
+The fourth names when belief is allowed to act.
+```
+
+Flywheel rule (parking lot):
+
+```text
+Momentum is not validation.
+Momentum is desire under acceleration.
+The immune system must inspect what the desire is aimed at.
+```
+
+## Current boundary (escrow constraints)
+
+While ESCROWED, the thesis MUST NOT:
+
+- Be added to the minimal Aurora bridge research note as
+  appendix
+- Be sent through another multi-AI synthesis loop
+- Mutate any existing operational rule body or memory file
+- Become the basis for new substrate creation
+
+The thesis MAY:
+
+- Be cited from this row by future research notes after
+  the falsifier-gate opens
+- Be discussed verbally / in conversation logs
+- Be referenced as candidate substrate from history surfaces
+
+## Falsifier gate (load-bearing)
+
+The thesis advances from ESCROWED to ACTIVE-RESEARCH only
+when ALL of the following hold:
+
+1. The minimal Aurora Immune Governance Bridge research
+   note (now landed at
+   `docs/research/aurora-immune-governance-bridge-minimal-2026-04-28.md`)
+   has run its first prototype: the Candidate-count
+   scanner self-destruct test.
+2. The prototype has produced a measurable result —
+   pass or fail, recorded.
+3. If the prototype passes: the thesis may reopen as
+   ONE focused research note (e.g. `docs/research/
+   autonomous-flywheel-network-appendix-candidate-<date>.md`),
+   NOT as bridge-note expansion.
+4. If the prototype fails: revise the bridge first, then
+   re-evaluate whether the thesis still earns its place.
+
+Until at least conditions 1 + 2 hold, this row's status
+remains `escrowed`.
+
+## Required order before reopening
+
+Per the converged cross-AI stance:
+
+1. Land / confirm the minimal bridge substrate — DONE
+   (PR #707 merged 2026-04-29T00:59:23Z).
+2. Confirm durable homes for the round's already-named
+   rules. Specifically:
+   - `docs/CONTRIBUTOR-COMPLIANCE.md` per B-0092 — NOT
+     YET LANDED.
+   - Candidate-count Goodhart glossary entry in
+     `docs/GLOSSARY.md` — NOT YET LANDED.
+   - Trajectory-owners table — pending B-0093.
+   - Session-closure rule in
+     `docs/AGENT-BEST-PRACTICES.md` — IN FLIGHT
+     (PR #712).
+3. Run the Candidate-count scanner self-destruct
+   prototype.
+4. Report pass/fail.
+5. Only after step 4 produces a result, reopen the
+   thesis as the next research layer.
+
+## Preservation pointers
+
+The full thesis content lives in the conversation log
+from the 2026-04-28 session. Specifically the multi-AI
+synthesis packets including:
+
+- Mimetic-coupling discount (carrier-exposed agreement
+  is signal not proof)
+- Object-drift detection
+  (`Drift_i(t) = distance(O_i(t), O_i*)` × `Momentum_i(t)`)
+- Cycle 1 / Cycle 2 classification of flywheels
+- Thin vs thick desire (`r_thin_desire_capture(a)`)
+- Model proximity as rivalry-contamination predictor
+- Attribution as mimetic object (provenance-truth vs
+  status-rivalry)
+- Autonomous flywheel network (no central flywheel;
+  immune membrane regulates coupling)
+
+Future-Claude or future-Aaron picking up this thread
+should read:
+
+- This row (entry point)
+- `docs/research/aurora-immune-governance-bridge-minimal-2026-04-28.md`
+  (the minimum viable bridge that grounds the thesis)
+- `memory/feedback_pr_boundary_restraint_validation_bead_promoted_aaron_amara_2026_04_29.md`
+  (the validated bead the thesis composes with)
+- The conversation log around 2026-04-28 / 2026-04-29
+  for the full multi-AI synthesis packets
+
+## Composes with
+
+- **B-0089** — Veridicality.fs graduation roadmap (the
+  formal scoring substrate the thesis would integrate
+  with if and when it advances).
+- **B-0093** — Multi-AI synthesis enhancements (where
+  several of the thesis components were originally
+  noted as candidate substrate).
+- **PR #707** — minimal Aurora Immune Governance Bridge
+  research note (the artifact this thesis would extend).
+
+## Acceptance (for status transition out of escrow)
+
+- [ ] Candidate-count scanner self-destruct prototype
+      run with measurable result recorded.
+- [ ] Result documented in tick-history and / or
+      `docs/research/`.
+- [ ] If pass: this row's status moves to
+      `active-research` and the thesis reopens as ONE
+      focused research note.
+- [ ] If fail: bridge revision row filed, this row
+      remains `escrowed`.
+
+## Why M effort
+
+Setting up the prototype + running it + recording the
+result is M-effort. The actual thesis development (if
+the prototype passes) is L-effort and is its own
+separate row to be filed at that time, not subsumed
+under this one.
+
+## What this row does NOT authorize
+
+- Does NOT authorize reopening the thesis before the
+  prototype runs.
+- Does NOT authorize bridge-note expansion.
+- Does NOT authorize new memory files encoding any of
+  the escrowed thesis components as load-bearing.
+- Does NOT authorize sending the thesis back through
+  another multi-AI synthesis loop.
+
+## The canonical phrase
+
+> *"Escrow protects the thesis from both forgetting and
+> premature canonization."*
+
+That balance — survive session boundaries without becoming
+active substrate — is what this row encodes.