Start here
Foundations
The objects and maps everything else builds on.
- 1Evaluation is measurementInteractiveA score is an observation, not a capability. Follow the instrument chain from construct to decision and the four places error gets in.10 min read
- 2The evaluator systemInteractiveMeet the master object E_sys = (C, I, R, V, L), and why a proof checker is a mechanism but a release gate built on it is a system.10 min read
- 3Vocabulary without category errorsReward model, rubric, PRM, RLHF, verifier: put each term on the axis it actually describes and most confusions disappear.10 min read
- 4The verification hierarchy, demotedInteractiveFormal > execution > learned judges > intrinsic signals. A useful prior on how grounded a signal is, never a verdict on validity.10 min read
- 5Marr's levels for evaluationBefore arguing about judges, find the level: what claim, what procedure, what machinery.6 min read