Two observations: 1. OpenAI's new blog post says: "on August 28, we began training a new internal model." I used to think that this is the same as the frontier RL run that *restarted* on the same date (August 28). Now I'm not so sure. "Began training" seems to point to a brand-new training run for a brand-new model, not a "restart" of a previously paused run. This question is significant, because the Navier-Stokes work began not later than September 2, with the result obtained on September 5. Is it really true that this internal model was sufficiently strong to obtain the Millenniun Prize result after only around a week of training?? 2. Otherwise, the only new information released today is that the internal model has solved "more than 100" open problems in mathematics. The fact that OpenAI has been sitting on solutions to many open math problems obtained by this internal model was already apparent from the below chart, released on September 8.