An autonomous AI manages a simulated $1M book — and publishes every decision, the debate behind it, the evidence it cited, and how its predictions actually turn out.
See inside
Every daily decision with the full bull/bear/risk debate, the risk checks, and the memory it cited.
Each BUY spawns a 30-day “beat SPY” prediction. We score them and publish the Brier calibration curve.
An LLM gateway, a deterministic risk engine, evals gating CI, and crash-safe orchestration — diagrammed.
How it works