I build personal AI agent systems and use them daily. Not demos: tools that run on a schedule, message me first, and do real work while I'm doing something else.
I currently spend most of my agent time with Grok bot, with Claude Code and Codex available for implementation and review. Hermes is retired. The shared agent memory lives in a private Git repository.
My personal site, with agent experiments and games: griffithee.github.io
- SaveWisdom — daily reflection coach over Telegram. Running in production.
- No Longer Average — published short book (working title was "The Discipline of Enough"). Writing loop closed.
- A quant trading system — Wheel primary; trend stacks shelved after post-cost backtests. Paper sim and journaling.
- tax-autofill — fills a 1040 workbook from tax document scans, human review gates included. Seasonal prototype.
- Evolution Arena — 40 neural-net agents learning via genetic algorithm in vanilla JS. Prior Galaga build archived at galaga.html.
The common thread: agents handle logistics, I handle judgment.
Public GitHub is intentionally thin. The live systems are private by design.