Essays on AI evaluation, constraint design, failure modes, and welfare measurement.https://themattnash.com/https://themattnash.com/writing/accuracy-is-the-floor/https://themattnash.com/writing/accuracy-is-the-floor/Why 93% accuracy still leaves the failure boundary unknown, and what calibration, constraint adherence, and consequential error rate actually measure.Tue, 07 Apr 2026 00:00:00 GMThttps://themattnash.com/writing/the-valence-gap/https://themattnash.com/writing/the-valence-gap/Current AI welfare frameworks are built to detect access consciousness, not valenced experience. That is the wrong target. Here is what a better framework would require.Tue, 31 Mar 2026 00:00:00 GMThttps://themattnash.com/writing/the-requirement-is-not-the-design/https://themattnash.com/writing/the-requirement-is-not-the-design/Every AI governance framework says human oversight is required for irreversible actions. That requirement, on its own, does nothing. Here is the difference between policy and architecture.Mon, 23 Mar 2026 00:00:00 GMThttps://themattnash.com/writing/evaluation-was-always-the-work/https://themattnash.com/writing/evaluation-was-always-the-work/Fifteen years in product leadership kept surfacing the same question: how do you know this system is ready to be trusted? That question scales. Here is why I am building toward it.Fri, 13 Mar 2026 00:00:00 GMT