Show HN: I turned my AI chat history into a portable cognitive fingerprint

1 point

3 months ago

I run two businesses across too many AI tools and got tired of re-explaining myself every time I switched contexts, so I built something that reads your conversation exports from ChatGPT, Claude, Gemini, or Grok and generates a structured model of how you actually reason. Not a summary of what you've talked about. Ten dimensions (incl. decision speed, abstraction level, conflict response) scored with Bayesian updating and 90-day temporal decay so the profile continually updates.

The fingerprint is portable via MCP. I launched it to be compatible with Claude Pro (Claude Code or regular .ai interfaces) on first message and it adapts. Onboarding is about 2 minutes: you upload an export, watch it build in real time, and get a before/after demo using a real question pulled from your own history. One generic Claude response next to one shaped by your fingerprint.

I built it solo with AI-native dev tools in about 30 days. pgvector semantic search, OpenAI embeddings, Claude Haiku for cognitive extraction, Supabase/Postgres backend, OAuth 2.0 with PKCE for Claude.ai, REST API with OpenAPI spec for ChatGPT.

First 25 beta users are free, $10/mo after that.

Try it: https://trypaolo.com

Happy to answer questions about how it works under the hood, and always welcome feedback/bug reports etc.