Scanner Lab

Operator hypothesis → Opus 4.7 generates scanner code → critic reviews → walk-forward backtest → activate.

AI-generated code runs in a sandboxed validator (allowlisted imports, no I/O, no eval, smoke-tested on synthetic bars) before any backtest. The backtest gate filters out anything below 20 trips / 40% win rate / +0.10R average. Operators review the artifacts (code, critique, metrics) before activating into the live roster. Every cycle costs ~$0.15–$0.30 in Anthropic tokens.

New proposal

Describe a market behavior you think has edge. Be specific about the SIDE, the SIGNAL (what triggers it), and the WHY (why it should work).

Hypothesis

Proposals

No proposals yet. Submit a hypothesis on the left to kick off the first pipeline.