I build agent operating systems and proof harnesses on real hardware —
then write the failure reports in public. If your mission needs someone who ships
on metal, deletes false README claims, and keeps the weird: this page is for you.
Who I am in a room
- High-agency solo systems builder — basement lab, 100+ dockerized experiments, local-first by default.
- Allergic to theater — I published the commit that removed features my README invented.
- Harness literacy — multi-agent loops, memory tiers, evals, budget routing; dual-Grok proof rails so models can’t fake disk writes.
- Creative stack — music, latent-space gallery, interactive demos. The brand is playful on purpose; the engineering is not a joke.
- Vertical operator — Iceboks AI Ops for SMBs; regulated product work (AuthZ/IDOR fixes that made it to merge requests).
Read these three first
Also useful
Good fit
- Open-source / open-weights labs and agent runtimes (Hermes-class, local models, evals)
- Startups that need player-coaches who build agents and the rails that catch bullshit
- Sovereign / self-hosted AI, edge stacks, weird technical product
- Causes that want a builder-writer, not a beige LinkedIn clone
Bad fit
- Roles that need me to hide the lab notebook and only ship slide decks
- SEO content farms and “AI blog automation” cosplay
- Teams that punish honesty about incomplete systems
Reach out
Tell me what you’re building and why it matters. Link a repo, a paper, or a broken production story.
I respond better to specific missions than to generic “we’re disrupting X.”
Aurora / Denver · lab on metal · iceboks