AI · AI today, November 6, 2026
Security
Independent evaluator finds Anthropic is now improving its AI twice as fast
· Aster editorial team
AI-generated imageMETR, a Berkeley nonprofit, reviewed about 11 weeks of Anthropic’s internal data and interviewed its staff. It concluded that the company is improving its models’ capabilities on well-specified tasks roughly twice as fast as a year ago. It ruled out more training compute as the cause; its best explanation is that AI is increasingly helping improve AI. Still, it says AI can’t yet train models on its own.
Why it matters
AI helping build the next AI could speed progress a lot, and also make it harder to oversee. This is one of the first times an outside group has measured it with access to a company’s own data.
For the technical crowd
The speedup is measured only on tasks that can be graded automatically, not on “messy” work. Anthropic redacted some information, and METR’s own team acknowledges uncertainty and internal disagreement. The report lands weeks before Anthropic’s IPO.
· Source: AP (vía ABC News)
More from AI today
- Companies use more AI, but spending has nearly stopped growing, Ramp saysRamp, a business-spending platform, reports that spending on “tokens” (the unit AI is billed in) rose just 1.5% between August and September, after…
- Amazon lets you shop by texting AlexaAmazon announced that customers can text-message Alexa’s AI agent to search for products, add them to a cart and buy them. You can also send a photo…
- Claude used to probe a 40-year-old cell mystery, with hits and missesAn Anthropic biology team used Claude to study vault particles, hollow barrel-shaped capsules first described in 1986 whose function is still…
- Anthropic pays for news for the first time in a deal with the APAnthropic signed its first licensing deal with a news organization, the Associated Press, according to Adweek. It let Claude show live AP election…
- Meta buys Conductor, a tool for running several coding agents at onceMeta acquired Conductor, a startup whose app lets developers run and supervise several AI coding agents at the same time. CEO Charlie Holtz confirmed…