Trades fire on technical-indicator thresholds — RSI-2, MACD, ATR, regime gates. Atomic bracket orders. NO LLM in the entry/exit loop. Rules are auditable, execution is reproducible.

7-layer institutional-grade: 1% max per trade, 2.2% portfolio heat, atomic brackets, correlation guards, drawdown circuit breakers. Same rules govern operator engine AND SaaS tenants.

9 Gemini agents continuously read macro (FRED + central-bank calendar), market (Alpaca + VIX regime), and business feeds (news + Reddit + earnings). Reflection agent proposes parameter refinements WITH evidence — operator approves.

Every new idea — operator-typed or AI-proposed — runs as a paper hypothesis first. Capital-bounded shadow trades validate edge against real markets before any promotion to live. Learn by trading without burning capital.

LLM-as-judge scores every decision against a 6-criterion rubric. Chatbot answers 'why did this trade win or lose?' using per-trade macro + sentiment + FOMC snapshots. Every winner AND every loser is explainable on demand.