The top things worth knowing about in AI today.
On 9 June Anthropic released Claude Fable 5, a new top-tier model that scores 80.3 percent on the SWE-Bench Pro coding benchmark, around 11 points ahead of the next model and well above its own Opus 4.8. It runs a 1-million-token context window and is free for Claude Pro, Max, Team and Enterprise users until 22 June, after which API pricing starts at 10 dollars per million input tokens. If your team writes or reviews code, it is worth testing against whatever you use now.
Read moreOpenAI confirmed on 8 June that it has filed a confidential S-1 with the SEC, the first formal step toward a public listing, though it says the timing is not yet set. The company is now booking more than 25 billion dollars in annualised revenue. The filing comes a week after Anthropic made its own confidential IPO filing, a sign the largest AI labs are preparing for public-market scrutiny of their spending.
Read moreJPMorgan has moved AI out of its experimental budget and into core infrastructure, alongside data centres and payment systems, inside a 19.8 billion dollar technology budget for 2026. Its internal LLM Suite is now used daily by more than 230,000 staff across over 500 production use cases, and the bank says AI has already returned about 2 billion dollars in savings. It is a clear sign that large enterprises now treat AI spending as non-negotiable rather than a pilot.
Read moreAlphabet launched an 80 billion dollar equity raise on 1 June to fund AI data centres, its first large stock sale since 2010, with Berkshire Hathaway committing 10 billion dollars. The company expects 2026 capital spending of 180 to 190 billion dollars, nearly double last year, and says demand for its AI services is outstripping supply. The scale shows how much capital the leading labs now need simply to keep pace with usage.
Read moreChina is drafting a plan to spend about 295 billion dollars over five years on a national network of AI data centres, Bloomberg reported on 9 June. State carriers China Mobile and China Telecom would run most of the sites, with at least 80 percent of chips and equipment sourced from domestic suppliers such as Huawei, largely shutting out Nvidia and AMD. It points to a self-sufficient Chinese AI supply chain separate from US hardware.
Read moreGeneralist AI raised 400 million dollars at a 2 billion dollar valuation in a round led by Radical Ventures, with Nvidia and Bezos Expeditions among the backers. The startup is building general-purpose foundation models for robots, software meant to control many different machines rather than one. The round is part of a wider bet that the next stage of AI moves from screens into physical work.
Read moreMicrosoft unveiled new in-house models on 2 June, including MAI-Code-1-Flash, which turns plain written descriptions into working source code. The aim is to lower costs for developers and reduce its dependence on OpenAI, still its main model partner. For Microsoft customers, it points to more model choice and cheaper options inside Copilot and Azure.
Read moreA White House executive order signed on 2 June states that federal policy should not require government licensing, pre-clearance or permits to build or release AI models. It also supports broad federal preemption of stricter state AI laws. For businesses, it signals a lighter-touch US approach that leaves more responsibility with the companies deploying AI.
Read moreA short read every week — the few things worth your time, and nothing that isn’t.