Open source maintainer building the OpenSRM ecosystem. Active GasTown contributor. Based in Dublin.
Thesis
A code review bot with 99.9% availability can approve terrible code dozens of times a day. Nobody’s tracking that.
Traditional SRE gave us SLOs for latency and availability. AI agents need SLOs for judgment quality. I’m building the tooling that connects these two worlds.
The ecosystem
Six independent components. Each solves a complete problem alone. Together they form a reliability lifecycle for AI systems.
Writing
GasTown contributions
Deacon plugin for the merge pipeline. Scores per-worker output quality, tracks trends, alerts on degradation. Merged by Steve Yegge. The Arbiter’s origin story.
Replaced tmux scraping with structured beads-based health detection. Merged via dual-model review.
Advisory
I help teams monitor AI agent decision quality in production.
If you’re running AI agents and asking how do we know they’re making good decisions? — that’s the problem I solve.
What you get:
Judgment quality monitoring integrated with your existing observability stack. Track which agents are reliable, spot quality degradation early, get alerts before bad decisions reach customers.
Fixed-scope setup: manifests written, monitoring configured, judgment SLOs measured, first accuracy dashboard delivered.
Get in touch ↗