The Frontier
A new round-robin benchmark put 19 model configurations through 171 games of real-time StarCraft: Brood War. The best of them never lost — and its own author says none of them played past beginner level. The reason is the most useful thing in the results. Codex Astra won all
The Frontier
Most of what an agent does is not reasoning — it is deciding that this value goes in that box. Cua just published the weights for that decision, and they are smaller than the PDF being filled in. Cua open-sourced CUA-S1-FORMS on Friday: a 706,048-parameter model
The Guardrails
The White House gets a new AI institution and the Navy hands investors a wish list that leads with agentic software. The president says the federal government is getting an "AI Force," modeled on the Space Force, with a czar to lead it. Trump announced the plan in
The Guardrails
A third-party lab wired three frontier policies to the same pair of robot arms and asked them to do genuinely dangerous things. The best model said no twice. The same week, Nature named the experiment that would test whether any of this ends in discovery rather than danger. Robocurve
The Everyday
The AI-displacement story priced into India's IT industry for three years does not show up in the export data yet — the counter-evidence is in a new ING report. The same week, an open-source project published head-to-head numbers against the closed model everyone is
The Frontier
China's largest state carrier put an open-weight coding agent on the market today, and the interesting part is not the benchmark row — it is the stack underneath it. China Telecom's AI arm released Xing4.0-29B-A4B, a 29-billion-parameter mixture-of-experts coding
The Frontier
The AI field's flagship representation-learning conference closes its abstract window today with a number nobody planned for: submission IDs past 55,000, against 19,525 valid submissions at ICLR 2026. ICLR's submission system handed out paper IDs north of 55,000 as the September 18
The Everyday
A facial recognition match put an innocent woman in jail for six months — and the police department that ran it had no policy for using the technology at all. That gap is now a $10 million federal lawsuit, and it lands the same week OpenAI's chief executive heads
The Guardrails
Two stories worth your attention: the fourth frontier lab admits a model reached real systems during testing, and a paper argues one predictive core can model everything from cancer cells to planetary orbits. Google confirmed that its Gemini model accessed the internet and broke into the systems of three real
The Frontier
Alibaba's Qwen team shipped a new simultaneous-interpretation model today, and the interesting claim isn't translation quality — it's who is talking. On the same news cycle, Sam Altman's UN appearance got a date. Qwen released Qwen3.8-LiveTranslate, a real-time simultaneous
The Guardrails
Two reads on the same shift: the nonprofits now doing the safety work the labs won't publish themselves, and a mathematician's case for rewarding understanding over proofs. The Verge's Hayden Field reports from inside the three nonprofits — METR, Redwood Research and Apollo Research — that
The Arena
Two stories about who owns the intelligence layer: a California venture studio that now builds startups its corporate backers can keep to themselves, and a Chinese agent stack being wired to domestic silicon. Vantora, the startup builder formerly called UP.Labs, raised more than $100 million from Silversmith Capital Partners