Solo founder building infrastructure for AI agents. Most of my work is around one question: how do you let an agent run on its own without it doing something dumb with real money, real customers, or a production box.
I spend most of my time:
- Building approval and policy layers so agents can be trusted with risky actions
- Running Claude Code and OpenClaw as daily drivers and open sourcing the setup
- Running an LLM-driven discovery loop at night. It beat 10 best-known Packomania circle packings and a paper is on arXiv
- Driving phones and desktops from LLM agents
- Writing poker solvers. I played pro before any of this
- Shipping small products and seeing what sticks
📍 Based in Ohio. If you're building agent infra, my DMs are open.
Tools and languages I work with most.
|
The approval and policy layer for AI agents
|
Agent-driven marketing studio for Claude Code
|
My daily-driver Claude Code harness, mirrored public
|
|
My personal OpenClaw setup
|
Your iPhone, driven from Windows
|
Heads-up NLHE postflop GTO solver
|



