Anthropic researchers have uncovered a hidden 'thinking space' within their AI model Claude, providing unprecedented insight into how large language models generate responses. Meanwhile, OpenAI has launched ChatGPT Work, a super app combining chatbot, coding, and other AI tools to transform workplace productivity.
- Anthropic identifies Claude's hidden 'J-space' mental process
- OpenAI releases ChatGPT Work super app with new GPT 5.6 models
- Advances highlight growing AI complexity and workplace integration
What happened
Anthropic developed a novel analytical tool named the Jacobian lens (J-lens) that allowed them to detect a previously hidden internal area called the 'J-space' within their large language model, Claude. This space contains related but unreleased words that reflect the AI’s ongoing processing before producing a final answer, akin to revealing the AI’s 'thoughts' behind its outputs.
In parallel, OpenAI unveiled ChatGPT Work, a new super app that merges several AI capabilities including its popular chatbot, advanced coding tools, and various newly introduced models. This app aims to function as an AI-enabled assistant for complex workplace tasks and arrived concurrently with the launch of OpenAI’s GPT 5.6 model series.
Why it matters
Anthropic’s discovery provides unprecedented transparency into how large language models like Claude internally represent and manipulate language concepts before generating their responses. This insight is key to understanding AI decision processes, potentially aiding efforts to improve model accuracy and safety.
OpenAI’s ChatGPT Work signals a major step toward embedding AI into everyday workflows by consolidating multiple AI functions into a unified interface designed specifically to collaborate with users on diverse work tasks, marking a maturation of AI from assistant to active collaborator.
What to watch next
Further research will likely explore how to leverage Anthropic’s J-space insights to refine AI interpretability and reduce risks associated with unpredictable AI behavior. Monitoring how these internal model states influence outputs across different tasks will be crucial.
The rollout of ChatGPT Work will reveal how businesses and professionals adopt AI super apps in real-world work environments. Observers should watch for integration with existing productivity tools, user feedback, and subsequent iterations combining AI models and features.