Larry Roberts

Generalist who ships. Comfortable across networks, self-hosted infra, and systems design; currently deep in Rust. More "make the whole thing work and prove it holds" than single-language specialist.

Rust, and the surprisingly deep problem of authorization for AI agents — tiered permissions, tamper-evident audit, human-in-the-loop approval for anything destructive.

AXIOM — the AI capability gateway: fail-closed policy, per-agent identities, human approval on destructive actions, hash-chained audit. Plus the adversarial suite that proves the wall holds.

Talking to anyone wiring AI into real infrastructure — homelab or production. Feedback on AXIOM, war stories about agents doing things they shouldn't, and how to keep humans in the loop. Say hey.