Anthropic looks inside Claude as OpenAI launches ChatGPT Work

Anthropic researchers built a tool called the Jacobian lens, or J-lens, to examine what happens inside Claude as it answers questions or completes tasks. The tool revealed a hidden area the company calls J-space, where words related to a response can appear even if they are not ultimately produced.

OpenAI unveiled ChatGPT Work, described as a long-awaited “super app” that blends its chatbot, coding tool and new models. The product is positioned to work both for users and with them, and arrived the same day as OpenAI’s GPT 5.6 models. OpenAI is also developing a fully automated researcher.

Other developments included humanoids performing teleoperated surgery on living animals by removing gallbladders from pigs, resuscitated human retinas responding to light 10 hours after death, and Meta beginning to charge developers for access to a new version of Muse Spark. A historical feature also traced reinforcement learning ideas back to 1943 experiments in which B.F. Skinner trained pigeons to guide missiles through trial and error.