Stability measurement suite: answer, reasoning, and counterfactual layers (spec + rung-0 implementation) - #159

Open

MaxGhenis wants to merge 7 commits into

mainfrom

claude/xenodochial-rosalind-362a7f

Commits on Aug 25, 2026

Add stability measurement spec (post-adversarial-review v1)

MaxGhenis
and
claude
committed
908456c
Layers 1+3 core: stability metrics, cache guard, counterfactual deltas (TDD)

MaxGhenis
and
claude
committed
398022d
Layer 2: reasoning-stability judge scaffolding (TDD)

MaxGhenis
and
claude
committed
d87ffdd
Stability suite CLI: stability-report, reasoning-stability, counterfactual-manifest/report, stability-cost-plan

MaxGhenis
and
claude
committed
fa477bd
Gold set, layer-2/3 runbook, rung-0 dry-run record

MaxGhenis
and
claude
committed
9b8221b
Refresh spec cost basis for the 32-model refreeze (grok-4.6, ox-alpha)

MaxGhenis
and
claude
committed
e372f34
Use CF_ID_SUFFIX constant in fingerprint join

MaxGhenis
and
claude
committed
9100cb6