chore: remove project-level memory, using ~/.claude/MEMORY.md instead

CLAUDE_MEMORY.md

@@ -1,68 +0,0 @@
-# Claude Session Memory
-
-This file persists across sessions. Claude reads this at the start of each conversation.
-
----
-
-## Project: Alpha Arena
-
-Autonomous AI trading system for Polymarket prediction markets.
-
-**Owner:** oho
-**Repo:** https://github.com/saymrwulf/alpha-arena
-
----
-
-## Session History
-
-### Session: 2025-02-06
-
-**What we did:**
-1. Completed backtesting enhancements (Monte Carlo, walk-forward, optimizer, comparison)
-2. Wrote THE_ALPHA_CHRONICLES.md — 25-page novel about using the system
-3. Deep research into Polymarket API integration
-4. **CRITICAL DISCOVERY:** The system is fundamentally broken for live trading:
-   - Token IDs truncated in LLM path (src/agent/llm.py:182)
-   - Web UI trading is commented out (src/web/app.py:301-302)
-   - Multi-agent coordinator has correct code but isn't wired to CLI/Web
-   - Silent error handling masks failures
-
-**Key files created:**
-- `docs/POLYMARKET_API_INTEGRATION_ANALYSIS.md`
-- `docs/CRITICAL_INTEGRATION_ANALYSIS.md`
-- `THE_ALPHA_CHRONICLES.md`
-- `src/backtest/walk_forward.py`
-- `src/backtest/monte_carlo.py`
-- `src/backtest/optimizer.py`
-- `src/backtest/comparison.py`
-- `src/core/resilience.py`
-
-**Status:** Prototype, NOT production-ready. Do not attempt live trading.
-
-**Next steps (if continuing):**
-- Fix token_id handling in LLM path
-- Wire multi-agent coordinator to CLI
-- Uncomment web UI trading loop
-- Add integration tests against real Polymarket
-
----
-
-## User Preferences
-
-- Prefers direct, honest analysis
-- Wants to understand "unknown unknowns"
-- Values depth over surface-level coverage
-
----
-
-## Instructions for Future Sessions
-
-At the start of each new session:
-1. Read this file first
-2. Check git log for recent changes
-3. Continue where we left off
-
-At the end of each session:
-1. Update this file with what was done
-2. Note any critical discoveries
-3. List next steps