SpeedTrader AI

SpeedTrader AI is an Alpaca-native autonomous trading intelligence system designed around a simple principle: AI can reason about a trade, challenge it, and veto it—but it cannot create trading authority. The system combines deterministic quantitative signal generation with options intelligence, adversarial AI review, portfolio-aware risk controls, cryptographically bound execution authorization, Alpaca MCP integration, reconciliation, decision journaling, reproducible backtesting, and decision replay. Rather than allowing an LLM to directly generate and execute orders, SpeedTrader constructs and validates deterministic trade candidates first. The AI operates as a subtractive critic: it can CONFIRM, ABSTAIN, or VETO a candidate, but it cannot bypass deterministic risk controls or authorize execution. Every execution must pass hard risk validation and a single-use authorization boundary. The options engine evaluates real Alpaca option-chain data, contract availability, liquidity, pricing, affordability, and defined maximum loss before a candidate can reach execution. The system is designed to fail closed around safety-critical components and uses reconciliation to resolve uncertain broker states instead of assuming an order succeeded. SpeedTrader also introduces deterministic decision fingerprints and replay. A historical decision can be reconstructed from its stored market state and deterministic pipeline, allowing the system to verify whether the same decision can be reproduced independently of the AI model. A built-in dashboard provides an operational view of portfolio state, decisions, risk, execution, system health, and research results. The project prioritizes reproducibility, safety, transparent decision-making, and honest validation over unsupported profitability claims. Paper trading is used for the competition environment, and backtesting is explicitly separated from claims about future performance.