Cleanroom
A Docker environment for running AI agent harnesses on chosen inputs only. The agent gets its own Linux workspace and never sees your home directory or existing project trees.
Look
Louder
Hank Beaver builds the systems that let software agents do real engineering work, and plays and records music at home. Handle everywhere: blasterpal.
L Rinput: scroll0 VU = -18 dBFS
Projects
Things I ship on my own time
Agent skills
Portable skills in the SKILL.md format. convo saves continuation briefs, hashout makes agents in different harnesses debate a decision, memorywipe audits loaded context, and tldr shortens prose.
Credits
Liner notes
Principal Engineer for AI software development at Tripleseat. I design the platform that turns agent-ready tickets into reviewed pull requests, and the practices around it: shared memory for agent sessions and context that works across tools.
Before that, Staff and Senior Engineer on the same team. My career is built on Ruby, Rails, SaaS, and AWS, and I stay close to the code while I teach it to other people.
- 2026 →
- Principal Engineer, AI SDLCTripleseat
- 2025
- Staff EngineerTripleseat
- 2023
- Senior EngineerTripleseat
- Stack
- Ruby, Rails, PostgreSQL, AWS, Docker
- Agents
- Claude Code, Codex, OpenCode, MCP servers, multi-agent workflows
- Instruments
- Guitar, bass, drums
- Coffee
- Ethiopian Guji