An AI-decision-provenance layer in Python/FastAPI exposing twelve Unlicense primitives, including a quant selection-bias rigor gate (Deflated Sharpe, PBO, walk-forward OOS, look-ahead audit in pure numpy) over real backtrader backtests, a provider-agnostic LLM backend factory, a Strategy Passport schema, and on-chain reasoning-trace anchoring via TracePublisher and ReasoningTraceRegistry.sol. An autonomous agent_runner reads vaults, rebalances via AMM, and anchors keccak256 traces through a real Circle DCW signer; 10+ Foundry contracts are deployed on Arc testnet with verifiable broadcast tx hashes, live at archimedes-arc.com.
5+ — Dan Browne Marten Windler Daniel Reis dos Santos Chuan Bai Önder Akkaya
AI trading strategies that fuse user intent with market data, academic research, and statistical rigor to help you win more than you lose