Running Trader Intelligence on my own trading data

I had years of closed trades on disk but no journaling — the patterns were locked inside CSV files and broker exports with no per-trade context or actionable learning.

Use my own product to classify every trade with Claude, see what the LLM thought I was doing right and wrong, and decide whether the output passed the bar to sell to other traders.

Built Trader Intelligence MVP. Trade ingest, Claude Haiku classification with prompt-cached system block, daily summary writer, Supabase persistence, public surface on noahduus.com that renders the run with real graphs. Ran it on my own data.

Claude flagged a recurring pattern I had not seen myself. The dogfood loop closed: I am running the product on my own data, the public surface shows real numbers, and the case study is now its own sales asset.