Writing
Thoughts on Responsible AI, software engineering history, and the gap between policy and practice.
I Spent an Afternoon Reading Model Cards. I Still Could Not Pick a Model.
Model cards have the right idea. But after an afternoon reading them I still could not answer the one question I needed ...
Your AI Governance Dashboard is Lying to You
Most AI governance metrics are designed to report upward. Nobody is building the metrics that help the team on the groun...
MAP Evolution: Why Your AI Risk Map Should Never Stop Changing
NIST says map your AI risks before deployment. Most teams do it once and move on. Here is why that is the same mistake w...
You Cannot Interview a Neural Network. The Hidden Cost of AI Decommissioning.
When I joined one company as an architect I discovered they had hired five of us to replace one person's ten years of in...
Governance That Lives in a Document is Optional. Governance That Lives in the Tool is Inevitable.
We had a retro agreement that developers would stop closing their own tickets. Everyone agreed. Nothing changed. Then I ...
We Don't Ask Developers to Be Their Own QA. Why Do We Ask AI to Govern Itself?
Two real AI failures. One from too much autonomy. One from oversight that existed but nobody could use. The fix is somet...