AI news without the empty calories.
The big stories. The useful context.
Everything worth your attention.
Follow the evidence ↗Carbon tops Artificial Analysis’s Intelligence Index v5. Initial access starts with selected cyber defenders, not a general release.
Opus 5.5 is being steered away from recreating a particular existing game, while original and user-owned projects are meant to keep working.
A research setup produced deliberately misaligned model organisms that passed a known monitor; Anthropic says the result is not evidence about deployed Claude.
The general training toggle may not cover Trusted Access cyber sessions; OpenAI has not yet documented the reported separate opt-out mechanics.
The December notice changes the accounting boundary, but OpenAI has not published the exact start date or token amounts.
The pricing row is official, but it does not establish general API access or universal eligibility.
The first in-orbit AI response is a commissioning milestone for a research prototype, not a deployed orbital data center.
What happened. Why it matters. Follow the evidence.
