What happened
In the same week as OpenAI’s Navier–Stokes claim and the Buckmaster–Alpöge forced-blowup papers, Adarsh Ganeshram, Valentin Duruisseaux, and Anima Anandkumar released two arXiv manuscripts (2609.10867, 2609.10860). Using a physics-informed neural network with a self-similar ansatz, they report an approximate singular Euler profile at the critical blowup rate 1/2, plus a framework that would prove nonlinear stability if a large finite set of constants can be certified. Tao’s 7 September note, updated after the preprint, treats this as a significant advance on the mainstream self-similar approach and as relatively AI-light: the PINN searched for a profile; literature review and partial Lean work were secondary. This is not Clay Navier–Stokes, not unforced Euler as a theorem, and not GPT-6 Astra or GPT-5.6 Sol. Credit stays with the documented PINN system so those named LLM releases are not given the search.