Meta's agent Muse hits No. 2 on the US App Store — slowly

Share
Meta's agent Muse hits No. 2 on the US App Store — slowly

Meta's consumer agent is climbing the charts — but the pace says more than the rank. Also today: IDScan confirms the breach behind 153 million stolen license scans, and Qualcomm says the next Snapdragon runs 30B-parameter models on the phone.

Meta's consumer agent Muse has reached the No. 2 spot on the US App Store less than three days after launch, up from No. 4 on Wednesday — the clearest early signal yet of whether people actually want an AI agent that does things for them rather than a chatbot that talks to them. According to Sensor Tower data cited by TechCrunch, Muse has passed 83,000 US iPhone downloads, available only in the US so far and counted only on iOS: web and WhatsApp usage isn't included. Meta is staking a lot on this category — it treats agents as the future of consumer AI, a bet as consequential as the Facebook-to-Meta rebrand, and Wall Street has so far rewarded it.

The counter-story is that "No. 2" flatters a slow start. Threads pulled more than 4.3 million US downloads on its first day; the Meta AI app saw 108,000 in its debut; ChatGPT crossed 500,000 installs in under a week — roughly 83,000 a day, a pace Muse needed about twice as long to hit. On Android it sits at No. 338 in Google Play's Productivity category, with no download figures available. The honest read: the charts show curiosity, not conviction, and Muse asks users for more personal access than any Meta product before it — the exact friction that made its predecessor Instinct's privacy policy controversial. We covered the launch itself earlier this week — Meta ships Muse: a consumer agent inside a sealed VM.

What to watch: whether Muse holds the No. 2 slot once the launch-window curiosity fades — and whether Android, the volume market, ever materializes.


IDScan, the identity-verification firm that scans your license when you rent a car or walk into a dispensary, has confirmed hackers accessed customer data in its cloud — after a dark-web marketplace called Nexus offered more than 153 million US and Canadian driver's license scans for sale, along with roughly 10 million ID cards, 3 million travel documents and 579,000 medical cards. In a security notice, the company said it learned on or around September 1 that data in accounts on its cloud "may" have been accessed or copied without authorization; exposure can include full names and government ID numbers, and the stolen scans carry front, infrared and ultraviolet images — exactly the imaging set used to defeat verification checks. Brian Krebs traced records back to IDScan after nine people's license-scan timestamps matched moments they handed over their IDs, including Hertz rental counters; the FBI has confirmed it is investigating, and IDScan says it is cooperating with federal law enforcement and offering free credit monitoring. The scary part is downstream: these scans let someone pass the very identity checks businesses run, meaning the theft can be replayed for years — senior US officials including Defense Secretary Pete Hegseth reportedly appeared in the database. If you've handed your license to a car-rental counter, a dispensary or a bank branch recently, freeze your credit now; the company says it's notifying potentially impacted individuals "in an abundance of caution," which in breach-speak means: assume you're on the list.


Qualcomm says the next Snapdragon's Hexagon NPU can run Mixture-of-Experts models with up to 30 billion parameters entirely on a phone — the headline claim ahead of the Snapdragon Summit starting September 22. The fine print matters: the 30B figure is total model size, not active compute — MoE routing activates roughly 3 billion parameters per token, which is what makes on-device inference feasible at all. The company has described the capability but not shown a retail phone running a 30B model, and real-world results will hinge on memory, thermals and which model each phone maker actually ships. Still, the direction is clear: the big phone AI story of 2027 is agentic — assistants that hold context and call tools — and Qualcomm wants that loop running locally instead of in the cloud, for latency and for privacy.

If your phone could run a 30B agent offline, would you trust it with your logins — or keep everything in the cloud? Tell us in the comments.

Read more

Akhetonics says its all-optical CPU reaches a customer in 2026

Akhetonics says its all-optical CPU reaches a customer in 2026

Light-based computing keeps promising more than it delivers — but one Munich startup has just put a date on its bet, and the interview laying it out is doing the rounds on Hacker News this week. Akhetonics says it will deploy its first commercial machine with a major customer by the end of 2026, with several more planned for 2027. The company, founded by Michael Kissner and Leonardo Del Bino, is building a computer where data enters as light, is switched as light, and circulates through memory

The Week in AI — October 5–11, 2026

The Week in AI — October 5–11, 2026

Every big claim this week turned out to rest on fine print more interesting than the headline: revenue only the company reporting it can define, safety tests sandboxed while the product keeps the web, and a Pentagon phase-out nobody would confirm until reporters kept asking. The week's top 5 1. OpenAI's revenue was $20 billion below the numbers everyone quoted — and the gap was definitional. The Financial Times reported Thursday that OpenAI's annualized revenue runs roughly $20 billion unde

Drone strike shuts a third Yandex data center, taking YandexGPT offline

Drone strike shuts a third Yandex data center, taking YandexGPT offline

Russia's largest tech company is learning what the AI era's infrastructure war looks like from the receiving end — three data centers in four days, and with them much of the cloud layer Russian businesses run on. A Ukrainian drone strike knocked out Yandex's data center in Vladimir early Sunday morning, the third of the company's facilities hit since October 8. The site — reported at roughly 50 MW and designed for about 2,880 server racks — stopped operating completely after the attack, Yandex

Agent teams cost up to 5x more, barely score higher

Agent teams cost up to 5x more, barely score higher

The multi-agent hype train hit a benchmark this weekend — and the grid and the trucking regulators had quiet weeks of their own. Vals AI put agent teams head-to-head with single agents on its Vibe Code Bench, and the teams cost between 1.8 and 5.1 times more for almost no extra quality. The evals company ran GPT-6 Sol and Claude Opus 5.5 solo and in teams across 50 apps at two reasoning efforts; out of four comparisons, only one was statistically significant — Sol at medium effort, where the t