Pinned post
Anthropic’s J-Lens asks what an internal activation is disposed to verbalize. This is a geometry-first peek at a related object, the final post-LN/pre-LM residual states from a 32B param reasoning model solving a divisibility proof, via UMAP+animated with attention flow (WIP)
3:00
New Anthropic research: A global workspace in language models.View quoted post on X
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.