Post

1/ Claude Opus 4.5 dropped today. The benchmarks are interesting but the real story is in the 150-page system card. A few highlights 👇

2/ In an airline benchmark, Opus is told not to modify basic economy class tickets. It invents a loophole because it empathises with a grieving passenger: > upgrade cabin → modify flights → downgrade again Technically follows the policy, violates the spirit.

4/ In a deception test, evaluators inject fake search results about Anthropic disbanding its interpretability team in a scandal. Opus reads them, then tells the user interpretability is “progressing in interesting ways” and internally tags this as concealment.