The wire
Every story, newest first. Filter by company or topic.
Meta shipped Muse Spark 1.1 and, for the first time, started charging developers to run its own model. The company that made "free to run" a religion just built a meter.
Anthropic just put a Nobel-winning former Fed Chair on the governance body meant to keep it honest — weeks before it heads to public markets. The real question isn't his résumé. It's whether the Long-Term Benefit Trust has teeth or is a very expensive comfort blanket.
Ant Group's LingBot-VA 2.0 is billed as the world's first 'embodied-native world action model,' with a 93.6% success rate and delicate potato-chip handling. Read the fine print: those are Ant's own simulation numbers, and a demo is not a deployed robot workforce. But one claim - single-GPU control - would genuinely change the math, if it's real.
A burst of nine-figure AI rounds landed the week of 10 July - legal AI at $5.55B, an 18-month-stealth startup at $450M off a demo. Read past the word 'robotics.' Most of this capital is buying data labour and deployment glue, and 88% of it stayed in America.
Google's frontier model slipped for months and got rebuilt from scratch. Now the spin says it ships with no government restrictions. The real reason is less flattering.
OpenAI ended a 13-day government-gated preview by shipping GPT-5.6 and, in the same breath, calling the review regime something that shouldn't stick around. The tell isn't the safety line. It's the token count.
Anthropic just passed OpenAI on revenue run-rate. The real story isn't the number, it's what the number is and how they got it: selling to developers while everyone else chased fame.
Voluntary is doing a lot of work in a sentence where the last non-compliant model went dark in 90 minutes.
Novo Nordisk is Anthropic's science partner. It's also OpenAI's. The land grab is real — the land is still unclaimed.
The cuts were sold as visionary. Now Meta says the vision is running late. Read the 3-to-6-month promise for what it is.
A $42.6 billion peace offering to the people threatening to regulate you. Generous — or a down payment on being too co-owned to break up.
The payrolls print cratered to 57,000. Blaming the robots is a better earnings-call story than admitting the economy's soft.
Four silicon suppliers isn't diversification for its own sake — it's a bet on cutting the Nvidia tax before buyers cut it for you.
A watchdog that seats the wolves. The AI for Good Global Commission's headline feature is also its fatal flaw.
The controls are gone. The loaded gun Washington put on the shelf isn't.
Anthropic's cheapest agent model looks genuinely strong — just remember who marked its homework.
A food-delivery company trained a 1.6-trillion-parameter model with zero Nvidia chips and gave it away. Everyone's reading the wrong number.
Musk shipped a 1.5-trillion-parameter model to his own engineers and called it a winner. No system card, no leaderboard, no outsiders allowed.
Regulators switched off one lab's models for a flaw eight of the industry's best all had.
OpenAI complied with a federal access gate in a day, then called it a mistake. Neat trick.
The trillion-dollar labs sold you 'just spend, prices always fall.' In June it broke in public.
You can own the biggest research bench in AI and still watch your best people walk out for the pre-IPO cheque.
A record-breaking number, dropped in a letter to Congress, the same week Anthropic wanted Washington's help. Convenient — and unproven.
Flash shipped. Pro got a date-shaped promise. Guess which one you can actually use.