01
Strategy collection
Five Algorithm Framework inputs compiled into typed specs and generated modules.
AI Algo Factory
Transmute strategy intent into validated trading systems.
An operator console for the Algorithm Framework pipeline: plain-language component inputs, AI extraction into typed specs, generated Python modules, auditable gate evidence, and explicit promotion controls before forward or live trading approaches the big red button.
Algorithm Framework
Five explicit components
Each framework plugin point gets its own natural-language input. The AI extracts typed specs, generates implementation modules, and surfaces validation issues instead of hiding trading assumptions in prose. Insights and portfolio targets are outputs that flow between these models, not separate inputs.
01
Universe Selection
02
Alpha
03
Portfolio Construction
04
Risk Management
05
Execution
Promotion gates
Evidence before execution
02
Feasibility testing
Prove the strategy executes trades and produces sane artifacts before heavier compute.
03
Walk-forward backtest
Purged walk-forward optimization through vectorbt.pro with fold-level evidence.
04
Robustness analysis
Monte Carlo and stress checks before promotion decisions get romantic.
05
Portfolio fit
Diversification and overlap analysis against the active strategy inventory.
06
Forward testing
Paper execution with reconciliation, alerts, and operator-visible evidence.
07
Live trading
Manual arming, kill switches, and audit trails — never automatic consent.
Current milestone
Engine foundation with operator surfaces
Convex persistence, registry workspaces, live OpenRouter draft generation, vectorbt.pro promotion diagnostics, and forward-test evidence dashboards are in place. Authentication and real broker adapters remain deferred until deployment posture is chosen.
My strategies
Live
Live progress, charts, and approve/repair/hold controls.
Builder
5×
Per-component prompts before spec extraction.
Live trading
Off
Paper evidence only; kill switches enforced.