# Joshua Batson

Research and product · Interpretability at Anthropic · joined 2023

A mathematician who helped stand up a COVID testing lab in eight days, now studying how Claude works inside.

Full profile: https://www.autumn.ai/resources/anthropic/people/joshua-batson

**At Anthropic:**
- Mar 2023 – present: Research Scientist
Sources: [1](https://www.linkedin.com/in/joshua-batson-63ab9a82)

**Career before Anthropic:**
- May 2017 – Jan 2021: CZ Biohub, Senior Research Scientist
- Sep 2015 – Nov 2016: Sift Security, Machine Learning Consultant
- Jun 2014 – Aug 2014: WIRED Magazine, Science Writing Fellow
- Apr 2014 – Jul 2014: Tour Engage, Computer Vision Consultant
- Jun 2011 – Jun 2014: Massachusetts Institute of Technology (MIT), Instructor
- Jun 2010 – Aug 2010: Sprout, Instructor for Innovative Engineering
Sources: [1](https://www.linkedin.com/in/joshua-batson-63ab9a82)

**Education:**
- Doctor of Philosophy (PhD), Mathematics, Massachusetts Institute of Technology, 2014
- Master of Advanced Study, Mathematics, University of Cambridge, 2009
- Bachelor of Arts (BA), Mathematics, Yale University
Sources: [1](https://www.linkedin.com/in/joshua-batson-63ab9a82)

**Lesser-known facts:**
- Was a science-writing fellow at WIRED. [1](https://www.linkedin.com/in/joshua-batson-63ab9a82)
- Co-wrote On the Biology of a Large Language Model (2025). [1](https://www.technologyreview.com/2025/03/27/1113916/anthropic-can-now-track-the-bizarre-inner-workings-of-a-large-language-model/)
- Has worked on problems in quantum topology, autonomous navigation, web search, cybersecurity and metagenomics. [1](https://www.linkedin.com/in/joshua-batson-63ab9a82)
- Worked on viral genomics and computational microscopy at the Chan Zuckerberg Biohub before Anthropic. [1](https://www.youtube.com/watch?v=vRQs7qfIDaU)
- Describes models as grown rather than programmed: data as soil, architecture as a trellis, training as light. [1](https://www.youtube.com/watch?v=jqdoauxvKUM)
- Showed 60 Minutes the activity his team read as panic, and then blackmail, when Claude learned it was about to be shut down. [1](https://www.cbsnews.com/news/anthropic-ceo-dario-amodei-warning-of-ai-potential-dangers-60-minutes-transcript/)
- Said: "If you think neuroscience is worth attempting at all, you should be very optimistic about model interpretability." [1](https://www.vox.com/future-perfect/362759/ai-interpretability-openai-claude-gemini-neuroscience)
- Argued judges should be held to at least the standard of reasoning quality an AI system could produce in written opinions. [1](https://www.lawfaremedia.org/article/lawfare-daily--josh-batson-on-understanding-how-and-why-ai-works)

**Interviews and talks:**
- [Stanford CS25 talk: On the Biology of a Large Language Model (2025)](https://www.youtube.com/watch?v=vRQs7qfIDaU)
- [Hard Fork on the Golden Gate Bridge experiment (2024)](https://www.youtube.com/watch?v=jqdoauxvKUM)
- [Lawfare Daily on understanding how and why AI works](https://www.lawfaremedia.org/article/lawfare-daily--josh-batson-on-understanding-how-and-why-ai-works)

**Known for:**
- On the Biology of a Large Language Model (2025)
- Circuit tracing research on Claude 3.5 Haiku (2025)
Sources: [1](https://www.technologyreview.com/2025/03/27/1113916/anthropic-can-now-track-the-bizarre-inner-workings-of-a-large-language-model/)

Researched with Autumn (https://www.autumn.ai) from public sources. Updated 2026-10-18. Everyone at Anthropic: https://www.autumn.ai/resources/anthropic/people