AI Trading Journey
This is the unfiltered story of building an autonomous AI trading system—every bug, every breakthrough, every lesson learned.
The goal: Reach $6K/month after-tax options income as fast as safely possible through disciplined, automated execution.
Strategy Profile: Automated iron condor system (SPY-first; expandable to SPX/XSP/QQQ/IWM) with 15-20 delta short strikes, $10-wide wings, and up to 5 concurrent positions.
Capital Deployment Policy: Alpaca paper account ($100K) is the primary validation engine. Alpaca brokerage account is traded opportunistically only when the same RAG + risk gates pass.
The method: A full autonomous stack where each layer has a strict role:
- Signal + market data layer: Alpaca + macro/news inputs feed the orchestrator.
- Decision layer: Claude Opus handles trade-critical reasoning; TARS/OpenRouter route non-critical tasks for cost control.
- Memory layer: LanceDB RAG retrieves prior failures and lessons before trade and code decisions.
- Execution layer: Orchestrator + trade gateway enforce a liquid index-options whitelist (SPY-first; expandable to SPX/XSP/QQQ/IWM), sizing, entry/exit, and stop-loss policies before orders hit Alpaca.
- Risk layer: Hard gates (position limits, drawdown controls, pre-trade smoke tests, mandatory checklists) block unsafe actions.
- Reliability layer: Ralph Mode and CI workflows continuously test, repair, and document the system.
Full architecture: Tech Stack.
Latest shipping timeline (reports + cross-post status): Beats.
Where We Are Today
Last Sync: 2026-02-25 22:05 UTC (source: data/system_state.json)
| What | Status |
|---|---|
| Account Equity | $100,845.25 |
| Daily P/L | -$48.72 |
| Win Rate | 100.0% (1 trades; target 80.0%) |
| Paper Phase | Day 14/90 |
| North Star Gate | ACTIVE (VALIDATING) |
| Open Positions | 1 structure(s), 4 option leg(s) |
| Unrealized P/L | +$155.00 |
Execution Focus: Do not scale risk until validation passes.
This section is auto-updated from data/system_state.json by scripts/update_docs_index.py via GitHub Actions.
Alpaca Account Snapshots (Auto)
Latest snapshots are refreshed by automation and published to GitHub Pages.
Paper ($100K) Snapshot + PaperBanana Financial Diagram
| Alpaca Snapshot | PaperBanana Diagram |
|---|---|
 |
 |
Captured: 2026-02-25T21:26:00Z
Technical readout: Paper Account: net liquidation value $100,845.25; daily P/L -48.72 (-4.8 bps) indicating a negative drift session; cumulative P/L +845.25 (+0.85%); low capital deployment at 1.8% utilization with cash $101,068.25; open position proxy 4; win-rate estimate 100.0% (n=1); North Star gate LOW.
Brokerage (Live) Snapshot + PaperBanana Financial Diagram
| Alpaca Snapshot | PaperBanana Diagram |
|---|---|
 |
 |
Captured: 2026-02-25T21:26:00Z
Technical readout: Brokerage Account: net liquidation value $211.02; daily P/L +0.00 (+0.0 bps) indicating a flat premium-decay session; cumulative P/L +191.02 (+955.10%); high capital deployment at 99.1% utilization with cash $4.00; open position proxy 0; win-rate estimate 0.0% (n=0); North Star gate LOW.
Progress Dashboard Snapshot
Captured: 2026-02-25T21:26:00Z
Manifest: /trading/data/alpaca_snapshots.json
Featured Posts
- The Silent 74 Days — How we built a system that did nothing for 74 days
- Complete Iron Condor Guide — Our full strategy and tech stack
- The Position Stacking Disaster — A bug that cost $1,472 in paper trading
Built by Igor Ganapolsky (CEO) with Claude-powered autonomous agents, including Ralph Mode.