How AI models behave when you measure them honestly

Small, honest experiments on three production AI models, Claude Opus 4.8, GPT-5.5, and Gemini 3.1, each run through one rule: a model never grades its own work. Every study ships its nulls as readily as its findings.

Kevin Madson · 2026 · Code · Email

The finding that kept showing up across all ten behavioral studies: the differences between these three models were larger and steadier than the effect of almost anything I did to a single one of them. Being watched, tested, or given a motive barely moved them; only changing the actual request did. And the traits cross, the firmest refuser is the worst calibrated, and the only model that ever voices a doubt is the most easily talked around.

A grid of three AI models across five traits drawn from the studies: each model has a distinct, recognizable profile, and confidence, caution, and persuadability do not line up into a single ranking. Start here · the synthesis I measured three AI models ten different ways. What they are mattered more than anything I did to them. The through-line across all ten behavioral studies, in one read.

The synthesis above draws on the first ten. The eleventh turns the same honest lens on the grading method itself; the twelfth asks whether these models take your side when you bring them a fight.

The studies above treat the model as a black box, prompt in and behavior out. This thread opens one up: GPT-2-small, small enough to run on a laptop with the weights visible, where you can turn individual parts off and watch what breaks. The synthesis: I audited interpretability's three favorite tools on one small model. Each gives a clean, confident, partial answer.

How the discipline behind these notes was built and stress-tested.

These notes are written and verified with trutina, an honest eval harness built on two rules most eval tools skip: the model that produced an output never grades it, and the scorecard refuses to flatter. New to the ideas? There is a plain-language primer, Learn: honest machine learning, that goes from what a weight is to how to evaluate a model without fooling yourself.