docs(research): status sync — E8 autopsy correction, E9 family, the lattice formalization

[drewstone]

Per the hold-out call on E8 — the autopsy validated it twice over:

E8: the corner is UNMEASURED, not negative

Ground truth (per-task artifact + a parser probe): the belief controller stopped after one shot on 14/16 unsolved tasks because its verdicts never existed — critique() routes through observe()'s findings-extraction protocol, which strips the VERDICT: format (the probe shows the analyst returning an ordinary recommendation under the verdict instruction). The reported 'trigger-happy STOP' narrative was wrong — never verified before narrating. Classification: infra/design-flaw. The next arm needs a verdict-capable channel (a raw firewalled analyst call on StrategyCtx) before the corner gets a verdict.

Also

• E9 status: three family runs recorded (HOLDs, the train-side displacements + reproducer certificate, the funnel-alignment law) — ledger pointers, results stay out of the doc.
• factorial-ablation-design gains the configuration-lattice formalization (the product space, J(c) vector response, fractional-factorial framing), the steering {0,1}⁴ with measured corners + the Hamming-ball reading, and the measured-cells update (σ×κ promoted ×2, honest billing).

All edits additive — designs and historical context preserved (the standing instruction).

[tangletools]

✅ Auto-approved PR — 79a6f420

Blanket team auto-approval is enabled for this reviewer service.
The full PR reviewer audit still runs separately and will publish findings if it detects issues.

_{tangletools · auto-approval · reason: blanket_auto_approve · 2026-06-11T01:45:54Z}