Data◆Article
Claude has a global-workspace-like internal structure — a small subset of activations called 'J-space' — that acts as a privileged representational substrate available for report, modulation, and flexible reasoning, distinct from simple chain-of-thought extraction.
Anthropic's interpretability research identifies 'J-space,' a small subset of Claude's internal activations that appears to function like a global workspace, usable for reasoning, reporting, and modulation. ✦ AI generated
Anthropic · Latent Space · 2026-07-07 · original ↗
Anthropic released research claiming a global-workspace-like internal structure in Claude, centered on a small subset of activations they call J-space. The core claim is not chain-of-thought extraction, but identification of a privileged internal representational substrate that appears available for report, modulation, and flexible reasoning.
Read full article ↗excerpt · fair-use quotation
- ·Anthropic finds global-workspace-like structure in Claude
- ·Centers on small subset of activations: 'J-space'
- ·Not chain-of-thought extraction, a deeper substrate
- ·Available for report, modulation, flexible reasoning
- ·Acts as privileged representational substrate
- ·Distinct from surface-level reasoning traces
- ·Supports both introspection and behavior modulation
Around this claim