Field Notes

Writing from the work.

Lessons from building AI that has to survive contact with compliance — economical, precise, and policy-safe, shipped as miniapps rather than loose agents. Longer than my daily diary, and written to be useful on its own. No email wall, no excerpt games.

June 12, 2026 · 8 min · AI governance

Human-in-the-Loop, On-the-Loop, Out-of-the-Loop — and Human-in-the-Lead

Four phrases people use interchangeably that mean four different things in production — plus a fifth that actually decides whether a system is safe to ship. A clear field guide to choosing the right AI autonomy level.