OpenAI used an unreleased model to produce a resolution to the Navier–Stokes existence and smoothness problem — one of mathematics' seven Millennium Prize Problems. Simon Willison flagged the result as significant, and Terence Tao has separately warned that AI is now "strip-mining" open problems at a pace that could make great research questions genuinely scarce.
---Security researchers released a demo of WeWorm, a zero-click exploit that spreads through WeChat calls on both iOS and Android. The victim doesn't need to answer or even touch their phone — the exploit succeeds silently. The bug was found with AI assistance.
---Automattic's board voted to put Matt Mullenweg on leave of absence. Mullenweg told staff via Slack that board members "conspired" behind his back — a remarkable fall for the man who built the infrastructure powering a third of the web.
---A new report details a DHS unit using predictive policing tools to pull people over — with minimal public disclosure. The story lands alongside news of a funk shirt engineered to fool AI surveillance systems.
---OpenAI's image generation has crossed 3 billion images across ChatGPT and the API. The new 2.5 release improves multi-turn instruction-following, speed, and style consistency — the model is getting genuinely harder to trip up.
---Mirendil founders — ex-Google and Anthropic researchers — are building AI systems that contribute meaningfully to their own development. The a16z conversation frames self-accelerating AI not as sci-fi but as the next logical engineering step.
Ethan Mollick declared he is "deskilling myself with AI so quickly on so many annoying tasks I don't want or need to be skilled at." From one of the most credible AI-in-education voices, this is a meaningful signal: intentional deskilling is becoming a legitimate productivity posture, not a warning label.
Miles Deutscher flagged that Anthropic researchers have put the probability of AI killing all humans within the next decade at under 10% — and treated that number as alarming rather than reassuring. When the people building the thing quote those odds and call it urgent, the rest of us should probably update accordingly.
Mollick noted that Google's absence from the frontier model race means missing the current run of AI-driven math breakthroughs — a pointed observation the day OpenAI announced a Millennium Prize solution. Falling behind on capability isn't just a product problem; it's now a science problem.
Two stories today deserve to sit next to each other: OpenAI's model solved a Millennium Prize Problem, and a separate research team used AI to find a zero-click exploit that silently compromises any phone that receives a WeChat call. Same underlying capability wave, opposite vectors. The math result gets the press release; the worm gets the security advisory.