AI Agents Prove Navier-Stokes Singularity
OpenAI mathematicians announced on September 8 that a swarm of roughly 10,000 autonomous AI agents, running on an internal advanced model, produced a Lean-verified proof of a singularity (“blow-up”) in the three-dimensional Navier-Stokes equations. The result addresses one of the six remaining Clay Millennium Prize Problems and shows that smooth solutions can develop infinite velocities in finite time under smooth forcing. A parallel effort by Tristan Buckmaster and collaborators using other AI models resolved related Euler and partial Navier-Stokes cases hours earlier.
The Navier-Stokes equations, formulated in the 19th century, govern viscous fluid flow and remain central to turbulence modeling. Prior analytic and computational work, notably by Córdoba and Martínez-Zoroa, established infinite-cascade techniques that produced singularities but with non-smooth forcing; the AI systems closed the remaining gap for the prize formulation. The proof was formalized in Lean after tens of hours of multi-agent interaction costing millions of dollars in compute.
Key uncertainties remain around human verification of the logical equivalence between the Lean statement and the classical problem, the elegance and insight of the AI-generated arguments (described by some as “slop”), and whether the idealized continuum singularity has direct physical implications given the molecular nature of real fluids. Priority disputes between the OpenAI and Buckmaster teams also highlight tensions in AI-assisted discovery.
Sources: Quanta Magazine, OpenAI announcements.
Hubble and Webb Reveal Primordial TNOs
On September 8–9, teams using NASA’s Hubble and James Webb Space Telescopes reported the discovery and characterization of 27 previously unknown trans-Neptunian objects (TNOs), including bodies as small as 5 km across—the faintest and smallest ever directly observed. The objects retain surface colors and compositions matching their much larger counterparts, indicating they have preserved chemical signatures from the solar system’s formation 4.5 billion years ago.
TNOs are icy remnants beyond Neptune that record early planetesimal conditions. Combined optical (Hubble) and near-infrared (Webb) observations of both “hot” (dynamically excited) and “cold” (stable-orbit) populations showed no systematic size-dependent color changes, contrary to expectations that collisions would alter small-body surfaces. Fewer small objects were found than some formation models predicted, and size distributions appear similar across populations.
Uncertainties include the precise mechanisms that allow surface preservation despite orbital scrambling and collisions, whether the lower-than-expected numbers require revisions to accretion models, and how representative the surveyed pencil-beam field is of the full Kuiper Belt. Follow-up spectroscopy and larger samples will be needed to resolve these points.
Sources: NASA, Phys.org, The Astronomical Journal.
Hidden Nickel Structure Accelerates Methane Conversion
Chinese researchers at the Dalian Institute of Chemical Physics reported on September 9 that an in-situ reconstructed atomic motif on nickel oxide is the true active site for partial oxidation of methane (POM), outperforming metallic nickel long assumed to drive the reaction. A low-loading (0.8 wt%) Ni/Al₂O₃ catalyst achieved 92% methane conversion and high CO/H₂ selectivity, matching higher-loaded catalysts while forming almost no detectable metallic Ni.
During reaction, the NiO surface reconstructs into a specific atomic arrangement that lowers the C–H activation barrier to 12.5 kcal mol⁻¹ (versus 38.5 for intact NiO and 15.7 for metallic Ni), according to DFT calculations. Operando observations were essential; conventional ex-situ analysis missed the dynamic structure.
Tensions center on scalability under industrial conditions, long-term stability of the reconstructed motif, and whether similar dynamic sites operate in other methane-conversion catalysts. The finding underscores the need for realistic operando methods over static models.
Sources: ScienceDaily, Nature Catalysis.