foxhunt

Files

jgrusewski 4938ac2ec5 plan(ml-alpha): v2 deployability validation — atomic-commit roadmap

Five-commit implementation plan + two-step runtime runbook for the
deployability spec committed at 07d5de504. Bite-sized TDD tasks:

  C1: wire save_checkpoint(best_h6000) into alpha_train.rs (~10 LOC)
  C2: max_drawdown_pct field on Summary, $35k starting-capital base
  C3: emit_deployability_verdict + tiered logic + 6 unit tests
  C4: GPU integration smoke test (#[ignore], env-var-gated)
  C5: three sweep YAMLs (smoke, threshold-tuning, deployability)
  C6: runtime — Argo prod training → smoke → threshold pre-reg → full sweep
  C7: commit verdict, update memory

Self-review confirms 1:1 spec section ↔ task coverage. Two soft
adaptation points (AlphaTrainSummary struct name, BacktestHarnessConfig
defaults) marked as code-read-and-adapt; no hard TBDs.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

2026-05-19 00:40:15 +02:00

infra

feat(infra): GPU training launcher with local/cloud routing

2026-02-23 11:19:58 +01:00

plans

design(foxhuntq): architectural pivot away from per-bar DQN to decoupled Belief Bus + Conformal DRL

2026-05-14 19:21:25 +02:00

superpowers

plan(ml-alpha): v2 deployability validation — atomic-commit roadmap

2026-05-19 00:40:15 +02:00

dqn-backward-nan-audit.md

fix(dqn): SP3 Mech 10 — ISV-driven h_s2 activation clamp + slot 49 diag