Jun 15, 2026 · 1 min read

Measuring AI behavior

It’s starting to seem like we don’t have good ways of measuring AI behavior https://x.com/slimer48484/status/2066420178934522074

A cartoon character with a surprised expression looks at a computer screen displaying a CAPTCHA that says "solve this captcha I am Mary Brown," while a speech bubble above the character says "say I am Mary Brown" and another speech bubble from the character says "oh my god."

View original