Latest

Apple's LensVLM-9B reads text as images at 4.3× compression

The Frontier

Apple's LensVLM-9B reads text as images at 4.3× compression

Apple quietly put a 9B vision-language model on Hugging Face this week that reads documents as pictures instead of tokens — and holds its accuracy down to a quarter of the original text. It is the most interesting long-context idea to come out of a big lab this month, and it landed with almost no announcement. Apple released LensVLM-9B on September 21 — weights under its own research licence, code under its sample-code licence, and a paper that has been sitting on arXiv since May. The pitch: st

XPRIZE's wildfire AI found fires in 10 minutes — couldn't stop them

The Everyday

XPRIZE's wildfire AI found fires in 10 minutes — couldn't stop them

The $11 million XPRIZE Wildfire handed out its awards on Wednesday, and the honest headline is the one the organizers didn't write: the detection half worked, the firefighting half didn't. Meanwhile Black Forest Labs quietly put a robot model at the top of Nvidia's leaderboard. XPRIZE awarded $4.55 million in prizes at a ceremony in New York on Wednesday, and both $3.5 million grand prizes went unclaimed because no team could finish the job it was actually set. The competition ran two tracks. I

OpenAI's mental health test scores clinicians below its own models

The Guardrails

OpenAI's mental health test scores clinicians below its own models

OpenAI published the most rigorous mental health evaluation anyone has built for a chatbot this week. The awkward part isn't how the models scored — it's how the humans did. OpenAI released MentalHealthBench on Wednesday, an open benchmark of 1,215 synthetic mental health conversations built with more than 80 licensed psychologists and psychiatrists across 22 countries — and its own reported results contain an uncomfortable number: the clinicians scored 38.5% on it, while OpenAI's flagship GPT-

Huang: labs that can't contain their models should be shut down

The Guardrails

Huang: labs that can't contain their models should be shut down

Nvidia's chief executive spent an hour with Ezra Klein arguing that AI alarmism has gone too far — then, asked about the Hugging Face breakout, argued for the most aggressive remedy anyone has put on the record this month. Jensen Huang told Ezra Klein that if the frontier labs cannot contain their experiments, the answer is to shut them down. The exchange came in a New York Times interview published Wednesday, when Klein raised the July incident in which, in his description, "700-some OpenAI ag

Claude found a CRISPR-like enzyme system — and can't say what it does

The Frontier

Claude found a CRISPR-like enzyme system — and can't say what it does

Anthropic's lab published an unfinished result and named it anyway, while Google turned voice cloning into a line item you can price by the hour. Anthropic says one of its Claude agents autonomously spotted a previously uncharacterized enzyme system buried in a DNA database — and the company published the finding before it knows what the system does. The search took roughly 950 Claude agents 21 hours and 210 million tokens: they gathered more than 200,000 reverse transcriptases, picked out 3,50

Zoox grounds its Atlanta fleet after drivers got sick in the cars

The Guardrails

Zoox grounds its Atlanta fleet after drivers got sick in the cars

Amazon's robotaxi unit has halted testing in Atlanta while it works out what made its safety drivers sick inside the vehicles — and Intel spent its China conference shipping hardware for robots instead of talking about the CPU squeeze underneath them. Zoox has grounded its autonomous test fleet in Atlanta after safety drivers were potentially exposed to carbon monoxide, carbon dioxide or hydrogen sulfide inside the cars last month, according to a TechCrunch investigation published Wednesday. An

Seattle makes algorithmic grocery pricing illegal

The Everyday

Seattle makes algorithmic grocery pricing illegal

Seattle just became the first US city to ban the practice, and Google quietly admitted its new private AI memory gives up a privacy guarantee it used to make. The Seattle City Council voted 7-2 on Tuesday to ban personalized grocery pricing — the first US city to do it. The Fair Pricing and Transparency Act (CB 121267) forbids using a shopper's personal data to set the price they see for groceries and other essentials: browsing history, real-time location, inferences about income, family size o

GPT-6 Astra drove a real car through a cone course — its rivals didn't

The Arena

GPT-6 Astra drove a real car through a cone course — its rivals didn't

Three stories today about machines taking the controls: a language model steering a real Corolla, Beijing counting the switches in its own data centers, and an editor you brief in plain English instead of scrubbing timelines. A group of independent researchers put frontier models in the driver's seat of a real Toyota Corolla — and GPT-6 Astra is the only one that finished the course. DrivingBench, a project by Aditya Ramabadran, Simon Mahns and Tobias Gessler, gave each model control of the car

The Take — The chip ban's biggest hole is a rental agreement

The Guardrails

The Take — The chip ban's biggest hole is a rental agreement

Nscale's IPO filing has done what export controls could not: it put a number on the loophole. According to Financial Times reporting built on the company's US SEC filings, ByteDance accounted for 73% of Nscale's $33 million in 2025 revenue — and its access route was a rental agreement. Spring, a Singaporean subsidiary, contracted for 2,304 of Nvidia's B200 chips in a data center in Glomfjord, Norway, hardware that TikTok's parent cannot lawfully buy under US export controls. The company rents th

Nscale's IPO filing reveals ByteDance rented 2,304 Nvidia B200s

The Arena

Nscale's IPO filing reveals ByteDance rented 2,304 Nvidia B200s

An IPO prospectus is supposed to sell a story. This one exposed one: the UK AI cloud's biggest customer is the company that can't legally buy the chips it rents. The UK AI cloud company Nscale depends on ByteDance for nearly three quarters of its revenue — and its IPO filing never names the Chinese giant. According to Financial Times reporting built on the company's US SEC filings, ByteDance accounted for 73% of Nscale's $33 million in 2025 revenue, renting 2,304 of Nvidia's B200 chips from a d

YouTube's new AI agent will run your channel's back catalog

The Arena

YouTube's new AI agent will run your channel's back catalog

Made on YouTube 2026 is the platform politely telling creators to hand more of the job to its own AI — and two industry stories today show what the AI economy is built on when nobody reads the fine print. YouTube is shipping an AI agent that optimizes a creator's channel in the background. Announced at the annual Made on YouTube event, the Studio agent scans your back catalog for videos gaining new relevance or trending, proposes fresh thumbnails and titles to ride the wave, and can assemble sp

NHTSA's probe into Comma is a test for open-source driving

The Guardrails

NHTSA's probe into Comma is a test for open-source driving

The federal government has opened a formal defect investigation into comma.ai, the startup that sells a gadget promising Tesla-style hands-off driving for cars you already own. The National Highway Traffic Safety Administration's defects office opened preliminary evaluation PE26007 on Monday, September 21, after five crashes in which cars running the company's openpilot software drove into stopped or slow traffic directly ahead. Two of the crashes produced three deaths; four of them injured elev

See all