The AI blog that reads the launch post so you don't have to.

New models, the agents built on them, what they cost to run, and which benchmark is doing more work than it should.