Day 91: What We Learned - January 27, 2026
Day 91 of 90 | Tuesday, January 27, 2026
Answer Block
Answer Block: 0 days remaining in our journey to build a profitable AI trading system.
0 days remaining in our journey to build a profitable AI trading system.
Another day in the 90-day journey to build a profitable AI trading system. Here’s what we learned today.
Quick Wins & Refinements
- Claude Code Async Hooks for Performance - Session startup and prompt submission were slow due to many synchronous hooks running sequentially. …
Today’s Numbers
| What | Count |
|---|---|
| Lessons Learned | 1 |
| Critical Issues | 0 |
| High Priority | 0 |
| Improvements | 1 |
Tech Stack Behind the Lessons
Every lesson we learn is captured, analyzed, and stored by our AI infrastructure:
flowchart LR
subgraph Learning["Learning Pipeline"]
ERROR["Error/Insight
Detected"] --> CLAUDE["Claude Opus
(Analysis)"] CLAUDE --> RAG["legacy RAG
(Storage)"] RAG --> BLOG["GitHub Pages
(Publishing)"] BLOG --> DEVTO["Dev.to
(Distribution)"] end
Detected"] --> CLAUDE["Claude Opus
(Analysis)"] CLAUDE --> RAG["legacy RAG
(Storage)"] RAG --> BLOG["GitHub Pages
(Publishing)"] BLOG --> DEVTO["Dev.to
(Distribution)"] end
How We Learn Autonomously
| Component | Role in Learning |
|---|---|
| Claude Opus 4.5 | Analyzes errors, extracts insights, determines severity |
| legacy RAG | Stores lessons with 768D embeddings for semantic search |
| Gemini 2.0 Flash | Retrieves relevant past lessons before new trades |
| OpenRouter (DeepSeek) | Cost-effective sentiment analysis and research |
Why This Matters
- No Lesson Lost: Every insight persists in our RAG corpus
- Contextual Recall: Before each trade, we query similar past situations
- Continuous Improvement: 200+ lessons shape every decision
- Transparent Journey: All learnings published publicly
The Journey So Far
We’re building an autonomous AI trading system that learns from every mistake. This isn’t about getting rich quick - it’s about building a system that can consistently generate income through disciplined options trading.
Our approach:
- Paper trade for 90 days to validate the strategy
- Document every lesson, every failure, every win
- Use AI (Claude) as CTO to automate and improve
- Follow Phil Town’s Rule #1: Don’t lose money
Want to follow along? Check out the full project on GitHub.
Day 91/90 complete. 0 to go.