Topic
Ongoing rivalry among leading AI labs, chiefly OpenAI and Anthropic, over model capabilities, pricing, and benchmark performance.
Anthropic pledged employee-level access for outside evaluators and OpenAI matched within hours. The two steps that ask labs to build less depend on rivals and governments agreeing a pace, and on taking each other's word for it.
Reality
Tristan Buckmaster says OpenAI accelerated onto Navier-Stokes after learning of his work with Anthropic's Levent Alpoge, and that the answer he got about his prompt history addressed lookup rather than training.
Reality
Zeki Data's flow numbers put Anthropic at 22-to-1 and Google DeepMind at 2-to-1. Staffers blame pay, morale and the Gemini mandate; Zeki's founder blames the publication rules.
Reality
Ramp's June card-spend index puts open-source and Chinese model serving in firms spending $248 per employee, 23 times the median. The cheap-model migration is starting at the top.
Reality
OpenAI's president has absorbed product strategy and the whole scaling arm while his title stayed the same. Altman is still CEO; the operating decisions are no longer his.
Reality
The company says expanded monitoring adds roughly 20 percent to the compute it covers, and that customers will not pay for it. It has not said how much of its compute is covered.
Reality
The gap is about 41 percent, the source is unnamed, and the company is preparing a share sale. Anyone signing a multi-year model contract should price the difference.
Reality