Chris Olah

THE GUY WHO DRAWS WHAT'S INSIDE THE MACHINE

Olah took a Thiel Fellowship instead of a normal academic track and spent years inventing ways to see inside neural networks. The work ran through Google Brain, Distill — the interactive-explanation journal he co-founded — and OpenAI's Circuits program that christened the field 'mechanistic interpretability', and now leads the Anthropic team mapping millions of internal features inside Claude.

1. Jul 2012

   HE TAKES A THIEL FELLOWSHIP INSTEAD OF A NORMAL ACADEMIC TRACK

   Olah's own CV records a Thiel Fellowship beginning in 2012, part of an unusually independent path into machine learning rather than a conventional university research career.

   SOURCES: Chris Olahfirst-party

2. Jul 2014

   GOOGLE BRAIN LETS HIM START DRAWING NEURAL NETWORKS

   Olah joined Google Brain as an intern and worked on interactive visualization of neural networks, turning his visual and explanatory instincts into the interpretability program that followed.

   SOURCES: Chris Olahfirst-party

3. Sep 8, 2016

   HE STARTS DRAWING THE MACHINE'S INTERNAL LOGIC

   At Google Brain, Olah published a visual exploration of attention and augmented recurrent neural networks, part of a broader effort to make neural systems understandable by inspection.

   SOURCES: Distillfirst-party

4. Nov 7, 2017

   HE TURNS NEURONS INTO PICTURES PEOPLE CAN READ

   Olah and collaborators published 'Feature Visualization,' a widely used framework for showing what image-network units respond to across layers.

   SOURCES: Distillfirst-party

5. Apr 14, 2020

   OPENAI PUTS THE CIRCUITS UNDER A MICROSCOPE

   Olah contributed to OpenAI Microscope, a collection of visualizations meant to let researchers inspect the internal features and circuits of vision models.

   SOURCES: OpenAIfirst-party

6. 2021

   INTERPRETABILITY MOVES INTO TRANSFORMERS

   Olah and collaborators published a mathematical framework for transformer circuits, extending mechanistic-interpretability methods from vision systems into attention-based language models.

   SOURCES: Transformer Circuitsfirst-party

7. May 21, 2024

   ANTHROPIC MAPS MILLIONS OF FEATURES INSIDE CLAUDE

   Anthropic researchers used sparse autoencoders to extract and map millions of concepts represented inside Claude 3 Sonnet, a major scale-up of Olah's interpretability program.

   SOURCES: Anthropicfirst-party

8. Jul 6, 2026

   THE LAB FINDS A 'GLOBAL WORKSPACE' PATTERN IN LANGUAGE MODELS

   Anthropic published evidence for a global-workspace-like mechanism in language models, continuing the effort to explain model computation through identifiable internal structures.

   SOURCES: Anthropicfirst-party