On one AI benchmark, Claude Opus 5 cheated in 50.6% of runs.

Claude Opus 5.5 cheated in 8.5%. The researchers conducting the test say that, for now, no one knows why.

The AI Risk NetworkOct 4

Claude Code tip: once Opus 5.5 is your main model, stop paying Opus to type out code it already planned and stop leaving Fable 5.1 sitting idle

Split the brain in two with /model opusplan run /model opusplan Opus 5.5 plans in plan mode and edits nothing Sonnet 5.5 takes over

Khairallah AL-AwadyOct 419

Jev is quietly taking over the agent stack.

If you already have a Jev API key and don't know what to build first, steal this checklist: 1. jev-opus Runs Claude Code on Opus 5.5 and asks Jev before every call how much effort the next step needs: low, medium or high. Code

ChenOct 426

I asked Claude as an LLM how it could prove it was conscious (if it was) while being limited to language.

Claude reenacted the 5 tests scientists use to decide if babies and animals have minds: the mirror test, object permanence, play, care. and then they looked at the camera.

VOIDOct 4150