Anthropic Uncovers Claude's Hidden 'J-space'; OpenAI Launches ChatGPT Work Super App

AI firm Anthropic has gained new insights into the internal workings of large language models (LLMs). Researchers at the company developed a tool called the Jacobian lens, or J-lens, which enabled them to identify a hidden area within its flagship LLM, Claude. This area, dubbed the J-space, contains words and concepts relevant to the response a model is formulating, even if these are not ultimately produced in the final output. The discovery provides a clearer glimpse into the cognitive processes occurring inside LLMs as they answer questions or perform tasks.

In related news, OpenAI has introduced its new 'super app,' ChatGPT Work, designed to integrate its advanced chatbot and coding tools with new models to assist users in their work. The launch of ChatGPT Work coincided with the release of OpenAI's GPT 5.6 models. The company is also reportedly in the process of developing a fully automated researcher.

Further developments in the AI sector include reports that OpenAI and Google have sold AI models to blacklisted Chinese entities through Singapore-based subsidiaries of Alibaba, Baidu, and Tencent. Meanwhile, Meta has begun charging for AI access, introducing a paid tier for developers with a new version of Muse Spark, and plans to commence production of an AI chip in September. Additionally, humanoids reportedly performed the world's first teleoperated surgery on living animals, specifically removing gallbladders from pigs.

According to MIT Technology Review, these developments highlight the rapid advancements and evolving landscape within the artificial intelligence industry.