Zac Hatfield-Dodds

Timeline

1. By 2019

   First-class honours at ANU

   In February 2019 he holds a degree in Interdisciplinary Studies with first-class honours from the Australian National University. His honours thesis describes Indigenous calendars and links traditional knowledge to weather observations to better characterise tropical seasonal patterns.

2. 2019

   Researcher at the 3A Institute

   In 2019 he is a researcher at the Australian National University's 3A Institute, building a new applied science around the management of artificial intelligence.

3. 2019

   Publishing the Hypothesis paper

   In 2019 he and David R. MacIver publish Hypothesis: A new approach to property-based testing in the Journal of Open Source Software. He is a maintainer of Hypothesis, Pytest, and Trio.

4. 2021

   A Python Software Foundation Fellow

   He becomes a Fellow of the Python Software Foundation in 2021.

5. December 2021

   A laboratory for alignment

   In December 2021 he coauthors A General Language Assistant as a Laboratory for Alignment with Amanda Askell and Yuntao Bai. The paper studies simple baseline techniques for a text-based assistant that is helpful, honest, and harmless.

6. 2022

   Alignment papers in 2022

   In April 2022 he coauthors Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback, on finetuning language models as helpful and harmless assistants. In July 2022 he coauthors Language Models (Mostly) Know What They Know, studying whether language models can evaluate the validity of their own claims. In December 2022 he coauthors Constitutional AI: Harmlessness from AI Feedback, which trains a harmless assistant from a list of rules or principles using RL from AI Feedback.

7. 2023

   A research team at Anthropic

   In 2023 he leads a research team at Anthropic and leads the Hypothesis project, while working towards a computer science PhD. He works at Anthropic, an AI safety and research company in San Francisco, and co-maintains Hypothesis and Pytest.

8. As of October 2026

   Technical staff at Anthropic

   He is a member of technical staff at Anthropic, an AI safety and research company in San Francisco, building reliable, interpretable, and steerable AI systems, and he is the lead developer of Hypothesis. In August 2026 he is a maintainer of Hypothesis, the property-based testing library used by roughly 5% of all Python developers, with David MacIver and Liam DeVoe.