Projects
Working code behind the talks. Everything here is public, and everything here runs.
The home lab these run on
BurritoBot Said Let's Dance, the Platform Said No
My latest talk, with Whitney Lee at KubeCon + CloudNativeCon North America 2026. The platform, not the prompt, is where the no has to live: identity, policy, egress and budget enforced below the agent. BurritoBot is the reusable Kubernetes platform behind it, the thing that gets set on fire on stage.
Site·Repo
A Hundred Agent Sessions on One Box
My KubeCon + CloudNativeCon North America 2026 tip talk. Coding agents belong on an always-on remote host, not on the laptop: close the lid, fly to Salt Lake City, open it again, and every agent is exactly where it was. The repo holds the config to borrow.
Site·Repo

Unleash an Agent, Watch It Burn
An AI-native internal developer platform, thirty-plus components, built so an agent can be turned loose on it. The demo behind the KubeCon workshop. What it teaches: what an agent does to real infrastructure when the only thing stopping it is the agent.
Repo

ArgoCD MCP Server
Give an AI agent real operational control of a production GitOps cluster without giving it the ability to quietly delete it. Fifteen tools across three safety tiers, read-only by default, 272 tests including integration tests against a real cluster.
Repo

Cloud Native AI Lab
Seven hands-on labs for running AI workloads on Kubernetes, including GPU scheduling.
Repo

Claude Dotfiles
The configuration layer I actually use: 70 skills, 15 agents, 21 rules.
Repo

Agentic Covenants
A practitioner framework for autonomous-agent governance. Six matrices mapped to NIST CSF 2.0, with working Kyverno policies, Falco rules, and kill-switch runbooks in every populated cell. The framework that tells a platform engineer which policy to write on Monday morning.
Site·Repo