EQ-01 — The best immediate action is not always the best policy
Scenario — equity_fire_sale_backfire (global_equity_ecosystem_v1, seed 42): a levered manager with a bearish thesis decides what to do with a large block.
MYOPIC — the greedy optimizer (per-round best response) chooses DUMP_LARGE_BLOCK. Value −0.655. Verdict: BACKFIRED — the market reads the forced sale, front-running and a price-pressure spiral follow.
FULL HORIZON — the Longstaff-Schwartz optimal-stopping policy, evaluated over 1,000 Monte-Carlo paths, chooses HOLD. Value −0.219, MC standard error 0.00713.
FORESIGHT GAP — +0.436, ≈ 66.6% of |greedy|. The first-action difference survives the increase from 100 to 1,000 paths while the reported uncertainty falls materially.
POLICY — POLICY_RECOMMENDATION: all 8 declared policy gates pass (report-only governance; the panel refuses to present the greedy verdict as "the optimal answer" and shows both side-by-side with the gap visible).
Kindynos exposes the optimizer disagreement instead of collapsing both analyses into one recommendation.
Boundary — a per-scenario result on the authored fire-sale calibration; no universal claim about HOLD.