Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-07-16 · high
The repo projects any layer's activation into output words, tested against Qwen and paired with a new paper on shared internal representations.
Why now: The repo posted July 2, and by July 16 it had already picked up 1,394 stars.
Anthropic released Jacobian Lens on GitHub July 2, code that decodes what a model's activations are nudging it to say, per the repo.
Interpretability researchers use tools like this to check whether a model's internal state matches what it later claims to be thinking. That's a live question for anyone building safety tools on top of a model's self-reports.
The method: average the Jacobian, the derivative of the model's final hidden state with respect to an earlier layer, across a corpus. That matrix becomes a linear map: feed it any layer's activation and it projects into the model's output vocabulary, skipping every layer in between.
It's built for open-weights decoder transformers. The formula is lens_l(h) = unembed(J_l @ h). The examples run on Qwen, not Claude, and the repo renders interactive layer-by-position visualizations of what an activation would say at each depth.
The release sits alongside a new paper, "Verbalizable Representations Form a Global Workspace in Language Models," on transformer-circuits.pub. Its claim: models route information through a shared internal format specific enough to decode this way. Jacobian Lens is the demonstration.
Anthropic marks the project explicitly as unmaintained, a reference implementation not accepting contributions, per the repo. That's an unusual thing to flag on a repo that already has 1,394 stars.
I read that label as the real news. Interpretability code built to prove one paper's point isn't the same thing as a library meant to run against your own model in production. Anthropic is telling you which one this is, before someone files an issue and waits for a merge that won't come.
Worth watching whether anyone reproduces the global workspace result on a model besides Qwen, since the repo doesn't include Claude itself.
Each link below shares sources, entities, or timing with this story.
Anthropic criticizes Qwen / Shared entities / Same source domain / Shared topic
Linked by a graph relationship (Anthropic criticizes Qwen); both cover Anthropic, Apache, July; reported by the same outlet (github.com).
Anthropic criticizes Qwen / Shared entities / Same source domain / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Anthropic criticizes Qwen); both cover Anthropic, July; reported by the same outlet (github.com).
Alibaba released Qwen / Shared entities / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (Alibaba released Qwen); both cover Apache, Qwen; reported by the same outlet (github.com).
Anthropic criticizes Qwen / Shared entities / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Anthropic criticizes Qwen); both cover Anthropic, July; overlapping topics (alongside, anthropic).
Linked by a graph relationship (Anthropic criticizes Qwen); both cover Anthropic, Qwen; overlapping topics (code, model).
Anthropic criticizes Qwen / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (Anthropic criticizes Qwen); both cover Anthropic, July; overlapping topics (alongside, anthropic, model).
Linked by a graph relationship (Anthropic criticizes Qwen); both cover Anthropic, Qwen; overlapping topics (anthropic, layer, model).
Anthropic criticizes Qwen / Shared entity: Anthropic / Same source domain / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Anthropic criticizes Qwen); both cover Anthropic; reported by the same outlet (github.com).