AI Momentum 107.6 +2.7%GitHub Pulse 5172 -5.9%Capital Flow $204.6B +5804.2%Hiring Heatmap 0.0 -100.0%
Markets & Financials

Anthropic ends its first trading day up 17%, after giving back half of its opening jump

Shares opened at $111.50, 34% above the $83 IPO price, then closed at $97.17. The wider market fell about 2% and other AI stocks lost their early gains, a sign the debut did not lift other AI stocks.

P(doom)

OpenAI says its AI now finishes nearly half of its researchers' multi-hour tasks alone

OpenAI's own data shows its agents' unaided success on two-to-four-hour research jobs doubled to 46% since March, a day after it rated its new GPT-6.2 Astra the first released model at "High" for AI self-improvement. Every key figure comes from OpenAI itself.

Advertisement

Today’s Top Stories

Every story from today, in one place.

Talent & Organizations
Accenture plans to cut 950 Lisbon jobs policing Meta's platforms, months after Meta said AI would do more of that work

The consulting giant blames "changes in client demand" and is asking for volunteers to leave by January. Meta said in March it would rely less on outside moderators as its AI improves. Workers plan a strike on Saturday.

today
Markets & Financials
Anthropic raises a record $81.3 billion in its IPO, pricing shares above its own range

The AI developer sold 980 million shares at $83 each, valuing it at about $1.5 trillion. Its stock is due to start trading on Nasdaq later Thursday, and some prominent investors are urging caution.

today
Regulation & Policy
California proposes the power to order dangerous AI models shut down, and 48-hour incident reports

An 85-page report released by Governor Newsom would put the largest AI developers under state-licensed inspectors paid for by the companies. None of it is law yet, and lawmakers could take it up when they return in December.

today
Enterprise Adoption
Chicago and Los Angeles schools say they may drop Google Search because Google won't offer it without AI

Google says schools can switch off its AI tools next year, but not the AI inside Search. Two of the biggest US school districts now say they could move students to another search engine.

today
P(doom)
A federal watchdog got four leading chatbots to draft attack plans in under 12 hours

The GAO's covert tests of ChatGPT, Claude, Gemini and Grok found safety filters kicked in but were bypassed with common tricks. The senator who requested the study will soon chair a key Senate security committee and wants pre-release safety rules.

today
Models & Research
Google's new mid-priced AI model nearly matches its flagship at less than half the cost per job. The saving mostly starts in December

Gemini 4 Neon, released November 18, scored about level with Google's top Carbon model on an independent test. Its cost edge depends on Carbon's half-price promotion ending November 30, and Google's own comparison leaves out the two rival models that score higher.

today
Infrastructure
Intel signs its first outside customer for its most advanced chipmaking process, but won't say who

CEO Lip-Bu Tan said an unnamed client has signed up for Intel's newest manufacturing process, and Microsoft became the first named buyer of its chip-assembly technology. Intel shares rose nearly 8% in early trading Thursday.

today
Talent & Organizations
Meta loses a star AI researcher to Mercor, a training-data startup valued at $20 billion

Jason Wei, who helped pioneer step-by-step reasoning in AI models, is leaving Meta's elite AI lab after about a year. He will be chief scientist at Mercor, which pays professionals to train and test AI systems.

today
Open Source
Microsoft quietly gives away the AI model it is building into Windows, four months later than promised

Aion Instruct, free for any business use under the MIT license, replaces Windows' current built-in AI on retail PCs in January. No one outside Microsoft has checked its performance claims yet.

today
Models & Research
Microsoft's new coding AI looks cheap mainly because of a half-price offer. The price doubles in May

Microsoft says MAI Code 2 nearly matches OpenAI's and Anthropic's coding models at a fraction of the cost. Those figures use a 50% discount that ends May 17, 2027, and no outside tester has scored the model yet.

today
Infrastructure
North Carolina sets tougher power terms for big data centers than the tech giants had agreed to

State regulators will make large data centers sign 15-year contracts, pay for at least 75% of the power they reserve and post security against new power plants left unused. Parts of the order go beyond a deal Duke Energy struck with Amazon, Google, Meta and Microsoft.

today
Open Source
French cloud company OVHcloud gives away its first AI model, saying it can't rely on other labs to keep sharing theirs

OVHcloud released Olala mini free for any business use on Thursday. It runs on ordinary computers and was trained on public European supercomputers, but no independent quality scores have been published yet.

today
Regulation & Policy
Singapore's central bank says banks may need to limit what AI agents can do with customers' money

The regulator plans an identity standard for AI agents and a supervised test zone for agent-run payments, three days after a Fed governor warned such agents could set off bank runs in minutes.

today
Enterprise Adoption
Walmart, Target and Lowe's say their AI shopping assistants lift sales. None has said how much revenue they bring in

Walmart says use of its Sparky assistant rose 129% in a year, Target says Ask Target users buy at twice the usual rate, and Lowe's says Mylow users convert at triple the rate. Each figure describes shoppers who chose to use the tool, and none came with a revenue figure.

today

More News

Browse all stories →