I wonder how this could be used to study creative thinking and the generation of new ideas. If I understand correctly, the J-space is some s…

I wonder how this could be used to study creative thinking and the generation of new ideas. If I understand correctly, the J-space is some sort of internal workspace where the model holds concepts to combines them and uses them for reasoning. You then could potentially inspect/intervene on that workspace during problem solving ask. Maybe inject an unexpected but structurally relevant concept into the model’s internal workspace and see if the model productively uses it in some way. Eg: solve a design problem but bring a totally unexpected reference to the thinking. Or a lateral thinking injection where you introduce ideas that are distant from the original ask but that have transferable structure, and see whether the model uses them to reframe the problem or generate more novel ideas. Also, the animation and video here is really well done and art directed. New Anthropic research: A global workspace in language models. Of everything happening in your brain right now, only a tiny fraction is consciously accessible—thoughts you can describe, hold in mind, and reason with. We found a strikingly similar divide inside Claude.

Source: Cristobal Valenzuela (X) | 2026-07-07