Anthropic has built a tool called the Jacobian lens (J-lens) and used it to uncover a hidden area - dubbed J-space - inside its flagship LLM, Claude. The J-space contains words related to the response the model is working on but may not ultimately produce. If Claude were a person (which it is not), you might say these hidden words reveal what's on its mind before it actually speaks. What they found ranges from the mundane to the unnerving.

Meanwhile, OpenAI has unveiled its long-awaited "super app" ChatGPT Work, which blends its chatbot, coding tool, and new models. It's designed to do your work for you and with you - and arrived the same day as OpenAI's GPT 5.6 models. The company is also developing a fully automated researcher, because apparently nothing says "progress" like machines doing all the thinking.

In other news: humanoids performed teleoperated surgery on living animals (removing gallbladders from pigs) in a world-first; SK Hynix landed the largest US listing by a foreign company, raising $26.5 billion; Tencent is leading a deal to unwind Meta's $2 billion Manus acquisition; resuscitated human retinas responded to light 10 hours after death; Meta has started charging for AI access; OpenAI and Google sold AI models to blacklisted China groups; a daughter tested an AI "death bot" of her father; an astronomer says the hunt for alien life needs more statistics; and Pokémon Go players turned Times Square into a giant battlefield, finally fulfilling the game's 2016 launch promise.

And in a blast from the past: In 1943, psychologist B.F. Skinner led a secret government project to teach pigeons to guide missiles by pecking at targets on a screen inside a warhead. The military never deployed Skinner's kamikaze pigeons, but decades later, those same principles helped power reinforcement learning - the technology behind some of today's most advanced AI systems. So next time you marvel at an AI's capabilities, remember: it all started with birds and snacks.