Software engineer in Denver @HashiCorp. Mostly Go and all security products. Work on Vault. Lead the launch of HCP Boundary.
Currently building secure runtimes and tools for agentic software: a language, an agent daemon, a training kit, and the guardrails that keep them honest.
Having a lot of fun doing some ML researching amateur style. Pre-training, fine tuning and LLM-as-judge experiments are my favorite right now.
type Guy struct {
Location string // Denver, CO
Writes []string // Go, Rust, TypeScript, Python
Runs string // The power bill up
}| lmkit | Train language models from scratch on one machine: pretrain, anneal, SFT, tokenizer, eval. |
| ago | A semantic edit protocol for Go. Agents submit compiler-checked mutations, so an edit that doesn't typecheck can't reach disk. |
| mlx-stack | A minimal LLM proxy and router for running multiple mlx instances. Supports text, vision, embedding and audio. |
| nevla | What happens when you vibe-code a language out of pure hatred for Python. Typed, error-handled, ML constructs first-class. |
| gyr | Agent daemon for local oversight. The gyrfalcon is the largest falcon. |
| gputex | A mutex for your GPU. Stops two jobs from fighting over one card. |
On GitHub since 2013. Still haven't run out of ideas for daemons.