Field notes

Field Notes

Notes from building Judith — a group agent — and reading the papers that matter: the scaffold around the model.

Agent securityJuly 2026

The Guard That Never Woke Up

My AI-security guard caught zero percent of the subtle prompt injections — not because it couldn't, but because it never invoked the detector. A continual-learning loop (RELAI) is what made the blind spot visible; the fix took base-split detection 0% → 94%.

Read →
Survey · ICML 2026July 2026

The Harness Is the Point

A cited, 45-paper survey of harness engineering at ICML 2026 — how the interesting design surface moved from the model to the runtime, the tool contract, the orchestrator, and the effect boundary.

Read →
Building JudithJuly 2026

How Do We Make It Realize Something Went Wrong?

Why closing the self-improvement loop is so much harder than drawing it — the mechanics are easy; the rater, the part that knows a wrong answer was wrong, is the whole game.

Read →