Side Quest
by contact142
"Multi-model collaborative problem-solving workflow with hardened backtest gates. Two rounds of brainstorm with a no-duplicate-thoughts rule between rounds, replay-not-stats backtests, walk-forward validation, sensitivity rejection, and per-quest verdict synthesis. Use when a hard problem stalls single-pass thinking, when you have testable data and want multiple model perspectives, or when an explicit creative-divergence step would surface non-obvious solutions. Works for any domain — software architecture, ML, infra, research design — anywhere replayable data exists."