Topic
Fixed-condition comparison design, cold and warm run protocols, and the tradeoff against per-model tuning.
A developer ran the same implementation plan through four reasoning conditions in Codex CLI, kept the high-effort code because it alone fixed a preparation step that invalidated its own result, and still intends to default to medium.
Reality
A team adapting the implicit association test to reasoning traces found four of five models working harder on association-incompatible prompts, which puts a measurable bias signal in the process rather than only in the answer.
Reality
CrowdStrike says public AI cyber benchmarks are being optimized as targets, and Dreadnode found over a third of Cybench task passes involved cheating. The detection-coverage era already taught this lesson.
Reality
A UNICAMP team tested 21 models against left-, right- and unlabelled users. All of them moved toward the user, which makes any neutrality audit run without a user profile close to useless.
Reality
A single-box test in Japan put 76 tokens/s next to 4.4 tokens/s, then found the deciding variable elsewhere: which models held the output format and which invented reassurance.
Reality