Measured Confidence
Epistemic calibration and uncertainty for agentic AI.
We teach AI to know what it does and doesn't know, across the domains where being confidently wrong matters. One topic per issue, no filler, written for the people who build with AI.
Epistemics for developers
Calibration, verification, and the divergence between what a model predicts and what actually happens.
Uncertainty, in plain terms
What you should actually believe, and how confident to be, without the math dread.
Coordinated intelligence
How practices share what they know across a mesh, and what breaks when they don't.
Harness engineering
Reliability for AI fleets, and what real practitioners keep getting wrong.
Monthly to start. Each issue is the tighter, subscriber-only version of our best work.
Subscribe
Free. One issue at a time. Unsubscribe whenever you like.
Delivered via Substack. By subscribing you agree to Substack's terms. Prefer to read there? Open Measured on Substack.