--- title: Anthropic Built a Tool to Read Claude’s Mind. Someone... description: Anthropic published research on July 6 introducing the 'Jacobian lens,' a method for tracing which internal activations in Claude shape its next token output,... canonical: https://daily.dev/posts/anthropic-built-a-tool-to-read-claude-s-mind-someone-else-used-it-to-rewrite-one--wad1gcmso twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Anthropic Built a Tool to Read Claude’s Mind. Someone Else Used It to Rewrite One. | daily.dev og:description: Anthropic published research on July 6 introducing the 'Jacobian lens,' a method for tracing which internal activations in Claude shape its next token output,... og:url: https://daily.dev/posts/anthropic-built-a-tool-to-read-claude-s-mind-someone-else-used-it-to-rewrite-one--wad1gcmso og:image: https://api.daily.dev/og/posts/Wad1gCMsO.png og:image:alt: Anthropic Built a Tool to Read Claude’s Mind. Someone Else Used It to Rewrite One. og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Anthropic Built a Tool to Read Claude’s Mind. Someone Else Used It to Rewrite One. **[Medium](https://daily.dev/sources/medium_js)** · 4 min read · 1 upvotes · 0 comments ## Summary Anthropic published research on July 6 introducing the 'Jacobian lens,' a method for tracing which internal activations in Claude shape its next token output, revealing a measurable subspace called J-space analogous to the global workspace theory in cognitive science. Within nine days, a solo developer built J-Wash (Jacobian-Brainwash), a local tool that turns this interpretability lens into a live model-editing tool. Users can load any Hugging Face model, observe which internal concepts fire per layer, and directly edit token directions — with changes previewing in real time and exportable as weight checkpoints. This approach sidesteps the cost of fine-tuning (no dataset, no training run, no GPU cluster) for tasks like adjusting tone, refusals, or persona. Caveats: J-Wash is two weeks old, built by one person, lacks an eval suite, and isn't production-ready. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is J-Wash and how does it let you edit a language model's behavior without fine-tuning? J-Wash, short for Jacobian-Brainwash, is a local tool built by a developer known as Extraltodeus that turns Anthropic's Jacobian lens interpretability method into an editing tool. It loads any Hugging Face model, shows which internal concepts fire per layer during chat, lets you pin concepts and edit their token directions directly, and previews changes live before exporting them into a real weight checkpoint (full model, specific layers, or LoRA) with no J-Wash code needed at inference. _Anyone weighing a fine-tuning run versus a lighter editing approach can track tools like this on daily.dev._ ### What did Anthropic's Jacobian lens research actually discover about Claude? Anthropic's paper, published July 6, introduced the Jacobian lens, a method for tracing which internal activations in Claude actually shape the next token, and used it to identify a measurable subspace called J-space that behaves like the global workspace theorized in human cognitive science. Researchers remain split on whether this proves anything about AI cognition versus simply describing a linear-algebra technique for tracing activation-to-token influence. _Developers following interpretability debates can keep up with findings like this through daily.dev._ --- Tags: [#llm](https://daily.dev/tags/llm), [#deep-learning](https://daily.dev/tags/deep-learning), [#anthropic](https://daily.dev/tags/anthropic) [View this post on daily.dev](https://daily.dev/posts/anthropic-built-a-tool-to-read-claude-s-mind-someone-else-used-it-to-rewrite-one--wad1gcmso) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Anthropic Built a Tool to Read Claude’s Mind. Someone Else Used It to Rewrite One.","url":"https://daily.dev/posts/anthropic-built-a-tool-to-read-claude-s-mind-someone-else-used-it-to-rewrite-one--wad1gcmso","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/anthropic-built-a-tool-to-read-claude-s-mind-someone-else-used-it-to-rewrite-one--wad1gcmso"},"datePublished":"2026-07-23T18:24:33.538Z","dateModified":"2026-09-14T06:13:39.762Z","description":"Anthropic published research on July 6 introducing the 'Jacobian lens,' a method for tracing which internal activations in Claude shape its next token output,...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/d6ab0a9a223ee01bc82d647f54feb2d2?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/d6ab0a9a223ee01bc82d647f54feb2d2?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Medium","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Medium","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/medium","url":"https://daily.dev/sources/medium_js"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/anthropic-built-a-tool-to-read-claude-s-mind-someone-else-used-it-to-rewrite-one--wad1gcmso","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,deep-learning,anthropic","timeRequired":"PT4M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Medium","item":"https://daily.dev/sources/medium_js"},{"@type":"ListItem","position":3,"name":"Anthropic Built a Tool to Read Claude’s Mind. Someone Else Used It to Rewrite One."}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/anthropic-built-a-tool-to-read-claude-s-mind-someone-else-used-it-to-rewrite-one--wad1gcmso#faq","mainEntity":[{"@type":"Question","name":"What is J-Wash and how does it let you edit a language model's behavior without fine-tuning?","acceptedAnswer":{"@type":"Answer","text":"J-Wash, short for Jacobian-Brainwash, is a local tool built by a developer known as Extraltodeus that turns Anthropic's Jacobian lens interpretability method into an editing tool. It loads any Hugging Face model, shows which internal concepts fire per layer during chat, lets you pin concepts and edit their token directions directly, and previews changes live before exporting them into a real weight checkpoint (full model, specific layers, or LoRA) with no J-Wash code needed at inference. Anyone weighing a fine-tuning run versus a lighter editing approach can track tools like this on daily.dev."}},{"@type":"Question","name":"What did Anthropic's Jacobian lens research actually discover about Claude?","acceptedAnswer":{"@type":"Answer","text":"Anthropic's paper, published July 6, introduced the Jacobian lens, a method for tracing which internal activations in Claude actually shape the next token, and used it to identify a measurable subspace called J-space that behaves like the global workspace theorized in human cognitive science. Researchers remain split on whether this proves anything about AI cognition versus simply describing a linear-algebra technique for tracing activation-to-token influence. Developers following interpretability debates can keep up with findings like this through daily.dev."}}]} ```