LLM thinking, deep research enhancements mid 2026 (Fable) are astonishing. Anthropic papers do partly explain the bases of such success (6 July 2026).

Global Workspace, J-space

Verbalizable Representations Form a Global Workspace in Language Models

(Picture generated via Gemini Pro based on text: Verbalizable Representations Form a Global Workspace in Language Models )

Added: Review of the paper by Neel Nanda