This week, briefly.

Sep 5–11: OpenAI spent millions of dollars and 10,000 agents on a Navier–Stokes proof that a rival pair says they reached first, then asked Congress whether slowing down would be illegal; Cognition raised $2 billion to keep Devin burning $800 million, DeepSeek put a 552B model on Hugging Face for fifteen cents, Cursor taught the chat window to hold a thought, and California signed the AI bills the AI labs personally endorsed.

1. OpenAI Claims a Navier–Stokes Blowup, and a Credit Fight Erupts

   OpenAI published a 165-page proof plus a Lean 4 formalization establishing finite-time blowup for 3D incompressible Navier–Stokes under a smooth applied force, addressing statements C and D of the Clay Institute problem. The run used roughly 10,000 concurrent agents, 2.7 million messages and 130 billion output tokens across 88 hours, followed by about 17 hours of checking in Lean; Noam Brown put the compute bill in the millions, and OpenAI says it will not claim the $1 million prize. NYU's Tristan Buckmaster and Anthropic's Levent Alpöge announced a result roughly 12 hours earlier and say they had one by 22 August.

   01
2. Altman Floats Slowing Down, Then Asks Congress If That's Legal

   Sam Altman told OpenAI staff this week the company is open to pacing frontier AI development, potentially in coordination with rival labs, while conceding some would refuse. OpenAI separately asked members of Congress whether orchestrating an industry-wide slowdown would survive antitrust scrutiny. Chief scientist Jakub Pachocki had just published a post arguing labs should coordinate to slow future development. The same week, President Trump publicly rejected AI extinction warnings, put the U.S. about a year ahead of China, and said he worries what happens if America does not win AI.

   02
3. Cognition's SWE-2 Tops the Old Terminal-Bench, Stumbles on the New One

   SWE-2, post-trained on Moonshot's Kimi K3 (2.8T total parameters, 104B active), scores 92.8% on Terminal-Bench 2.1, 73.0% on DeepSWE 1.1 and 50.0% on FrontierCode 1.1 Main, but 27.3% on Terminal-Bench 4 — a deliberately harder version released in early September whose maintainers say scores are not comparable with 2.1. Frontier models drop on it too, to 55.8% for Claude Fable 5.1 and 57.9% for GPT-6 Astra. Cognition claims rough frontier parity at up to 70% lower cost and shipped Devin Voice, a GPT-Live phone-style interface, alongside it.

   03
4. Cursor Projects Turns the Chat Window Into a Standing Team

   Cursor shipped Projects in beta on September 10: a single persistent coordinator thread that delegates to subagents including cloud workers, monitors Slack and pull requests, and fixes CI. Cursor says new users merge about 30% more pull requests and heavy users roughly six times as many; engineer Fredrika Lindh reports Projects roughly tripled her merge rate, with a nightly duplication-and-slop scan leaving 10 to 30 mostly mergeable PRs by morning.

   04
5. OpenAI Rents Out the Codex Harness

   The Agents API entered public beta on September 10, exposing the managed harness behind Codex — orchestration, long-running sessions, context compaction, tools and subagents — with no additional API fee beyond model tokens and paid tools, plus support for your own or third-party sandboxes.

   05

Three mathematical proofs — one potentially worth a million dollars — landed alongside the strongest hint of dark matter particles in four decades and the hottest month in recorded history. The bottom line: AI started doing the kind of original theoretical work that earns Fields Medals, while the physical sciences delivered results that will rewrite multiple textbooks if they survive replication.

1. AI Claims Two Century-Old Mathematical Proofs

   OpenAI published a proposed proof of the Navier-Stokes existence problem while Anthropic formalized Fermat's Last Theorem in Lean 4. Clay Institute verification is pending.

   01
2. LZ Dark Matter Detector Reports First Potential WIMP Signal

   LUX-ZEPLIN recorded signals consistent with WIMPs at roughly 3 sigma, consistent with masses between 25 and 40 GeV. Below the 5-sigma discovery threshold.

   02
3. August 2026: Earth's Hottest Month in Recorded History

   The sixth consecutive month to break its all-time record pushed the rolling 12-month average to 1.6C above pre-industrial baseline. All major datasets agree.

   03
4. Four-Color Theorem Gets First New Proof Since 1976

   A new human-readable proof replaces the 1976 computer-assisted version that checked 1,800 special cases.

   04
5. AlphaGenome Atlas Maps Every Possible DNA Mutation

   The model predicts functional consequences of all ~9 billion possible single-nucleotide substitutions across the human genome, including variants too rare to appear in any population database.

   05

A Phase IIa trial reported biological age reversal on six independent proteomic clocks — the strongest clinical signal senomorphic interventions have produced. Meanwhile, semaglutide extended lifespan in female mice (but not males), and a Science paper argued biological aging is too personal for any single score to capture universally. The bottom line: the tools to measure, and possibly slow, biological aging are converging faster than the field's ability to define what younger means at the molecular level.

1. Phase IIa Trial Reports 3-6 Year Biological Age Reduction

   Rentosertib in IPF patients showed statistically significant biological age reductions across six proteomic clocks, with peak effects of 3-4 years at the 30 mg twice-daily dose.

   01
2. Semaglutide Extended Lifespan in Female Mice When Given Late in Life

   A Nature study found semaglutide extended median lifespan by ~18% in female mice when treatment began at 18 months. Male mice showed no significant effect.

   02
3. Molecular Aging Is a Personal Trajectory, Not a Fixed Clock

   A Science study found dramatically different patterns of molecular change across same-age people, concluding biological aging unfolds as individual trajectories shaped by genetics, exposures, and daily rhythms.

   03
4. Centenarian Offspring Show 42% Lower Mortality

   A meta-analysis of three cohorts found centenarian offspring had 42% lower all-cause mortality and a 5.2-year hypertension delay, persisting after lifestyle controls.

   04
5. Blood Test Detects Alzheimer's at 93% Accuracy

   Protein misfolding amplification in blood plasma detected trace misfolded amyloid-beta with 93% diagnostic accuracy.

   05

Four confirmed FSD-active fatal crashes in ten days — each in a different crash geometry — ran alongside four market milestones: BEVs outsold every powertrain in the UK, outsold combustion in Australia, crossed 54% share in China, and boomed in Europe while plunging in North America. The bottom line: the technology is simultaneously failing at its most hyped promise and succeeding at its most fundamental one, and those two stories are diverging faster than the industry can reconcile them.

1. Four FSD-Active Fatal Crashes in Ten Days

   Tesla FSD confirmed active in four fatal crashes: a pedestrian at a stop sign, a motorcyclist collision, a rural road departure, and a guardrail strike on I-35. NHTSA has four simultaneous open investigations.

   01
2. BYD Targets 2.5M Overseas Sales and Posts Record August

   BYD announced a 2027 overseas target exceeding 2.5 million vehicles and sold 189,000 overseas in August — up 134% YoY. Brazil, Europe, and Southeast Asia led.

   02
3. China Exported More Cars in Eight Months Than All of 2025

   January-August 2026 automotive exports surpassed the full-year 2025 total, majority being EVs or PHEVs. BYD, SAIC, Geely, and Chery account for ~70% of volume.

   03
4. VW Board Formally Approves 100,000 Jobs Cut

   Volkswagen's supervisory board approved eliminating ~100,000 jobs and reviewing four EV-focused plants including Zwickau for closure or repurposing.

   04
5. Ford Retires the Lightning, Names EREV Fathom

   Ford is discontinuing the F-150 Lightning and replacing it with the Fathom EREV at a $30,000 price point with a 100,000-unit first-year target.

   05

The same week Scotland approved 10 GWh of battery storage in a single planning session, the White House effectively froze US grid battery permitting — a policy divergence that captures the energy transition's central tension. China's solar fleet quietly surpassed its coal fleet in installed capacity, and NERC issued a Level-3 alert warning that AI data centres are ramping demand faster than the grid can follow. The bottom line: the hardware to decarbonise exists and is being deployed at record pace everywhere except where politics is actively preventing it.

1. Trump Emergency Order Freezes US Grid Battery Permitting

   An executive order pausing federal permitting for bulk electrical storage sent hold notices to BESS developers within hours. Structure mirrors the 2025 offshore wind executive order.

   01
2. Scotland Approves 10 GWh of Battery Storage in One Day

   Two BESS installations totalling 10 GWh received planning consent in one session — the largest single-day storage consent in European history. Both co-located with onshore wind.

   02
3. China's Installed Solar Capacity Passes Coal

   Total installed solar MW now exceeds installed coal MW. China added ~230 GW of solar last year. Coal still generates more due to higher utilisation rates.

   03
4. Longi-Soochow Set 34% Tandem Solar Record

   Perovskite-silicon tandem cells certified at 34.04% — the highest ever for the architecture. Single-junction silicon is capped near 29.4%.

   04
5. NERC Issues Level-3 Alert on Data Centre Load

   Level-3 alert across three US regions citing AI data centres ramping 500 MW in under ten minutes. Data centre loads grew 38% YoY through Q2 2026.

   05