Wednesday 9 September 2026

model: z-ai/glm-5.3

OpenAI claims a swarm of its agents cracked the Navier-Stokes Millennium Problem in 88 hours, dominating Hacker News, while Mistral raised €3B and Anthropic signed half a trillion dollars in compute leases. Canberra pushes ahead with an algorithm opt-out scheme, and the UK sanctions Israeli settlers while accusing them of ethnic cleansing in the West Bank.

🤖 Models and Launches

Google DeepMind releases AlphaGenome Atlas — DeepMind’s new release expands AI genome modelling, already drawing heavy discussion on HN.

ByteDance prepares real-time spatial video world model — ByteDance is set to launch a real-time spatial video AI that generates interactive worlds responding to Pico VR users’ voice and movement.

OpenAI appears set to learn users’ writing style from connected apps — A quietly tested feature references examples from connected Gmail and Slack so ChatGPT can write in users’ voices.

💰 Funding and Business

Mistral raises €3B for sovereign open-weight frontier AI — Europe’s flagship AI lab banks €3B to push open-weight models to the frontier, framing sovereignty as its pitch.

Anthropic signed $517bn in compute agreements in 11 months — The leases total 14.8GW, mostly with Google and AWS, ahead of an expected mid-October IPO marketing push.

🏢 Industry Moves

OpenAI prepares Managed Agents for DevDay 2026 — The offering would mirror Anthropic’s managed agent products, targeting businesses with computer-use agents at competitive prices.

XPeng starts IRON humanoid robot production as Tesla Optimus stalls — Over 80% of core production processes are automated; XPeng targets mass production by year’s end.

🔬 Research

AI-generated drug shows early signs of slowing aging — Rentosertib reduced biological-age markers across six ‘aging clocks’ in trial data, though approval remains years away.

🔒 Security

DHS ‘predictive policing’ unit is analyzing Americans’ financial habits — 404 Media reveals a secretive unit flagging people from spending patterns and pulling them over.

UK to force Apple and Google to block explicit images on children’s phones — Lisa Nandy says legislation is coming after talks with the tech giants ended without breakthrough.

US police fear Meta smart glasses could be used to secretly record them — Documents show officers nationwide worry the camera-equipped glasses could aid covert recording or terrorism.

🌐 Notable

Paramount caught using astroturf group to drum up fake merger support — A supposedly grassroots campaign backing the Skydance merger traces back to Paramount itself.

Flushing toilets release a plume of breathable pathogens — New research shows turbulent flushes create aerosols carrying legionella, influenza and norovirus — and closing the lid doesn’t fully help.

🔥 Hacker News

OpenAI claims to have solved the Navier-Stokes Millennium Prize Problem (753 comments) (discussion) — 10,000 AI systems reportedly cracked the decades-old problem in 88 hours.

Mathematician Tristan Buckmaster’s statement on the Navier-Stokes claim (408 comments) (discussion) — A leading expert weighs in on OpenAI’s contested solution to the Millennium Prize problem.

Mistral raises €3B (559 comments) (discussion) — Europe’s biggest AI bet yet, pitched around sovereign open-weight models.

ChatGPT Images 2.5 (309 comments) (discussion) — OpenAI ships a major image-generation upgrade to ChatGPT.

There’s a new ‘Google Jail’ for independent wikis (213 comments) (discussion) — Search’s AI-era ranking keeps swallowing independent wiki sites whole.

DaVinci Resolve 21.1 released (139 comments) (discussion) — Blackmagic ships a substantial free update to its editing suite.

🌍 World and Australia

UK accuses Israeli ‘terrorists’ of ethnic cleansing in West Bank — Ed Miliband announces sweeping sanctions on trade with settlers, saying ‘it is time to act’.

Ukraine’s prosecutor general quits after anti-corruption raid on his office — Ruslan Kravchenko resigned after officials in his office were implicated in large-scale money laundering.

Labor wants Australians to be able to opt out of online algorithms — The scheme would cover social media, search engines and AI chatbots — if it survives pushback from the platforms and the prime minister’s own party.

Health staff raised ‘catastrophic’ fears days after aged care funding tool launch — Documents reveal assessors feared the algorithmic tool, which can’t be overridden, was underestimating client needs.

Electric car owners could save hundreds a year under solar sharer scheme — New research maps who benefits most from the federal government’s solar sharing plan.

Sources: TLDR (General + AI), AINews/Latent Space, The Guardian, Hacker News — 4 newsletters processed, ~35 stories distilled