Thomas Broadley

About me

I'm a Canadian living in Berkeley, California.

I work as a Member of Technical Staff at METR. METR studies AI capabilities, including broad autonomous capabilities and the ability of AI systems to conduct AI research and development. Our goal is to help society understand the capabilities of frontier AI systems, and what risks they pose.

Recently, I've been working on a team that's developing monitorability evaluations. We published some of our work in this blog post.

I'm also a co-author of three of METR's papers:

• "Measuring AI Ability to Complete Long Tasks", in which we demonstrate that AI performance on certain software tasks is doubling every seven months
• "HCAST: Human-Calibrated Autonomy Software Tasks", a software development benchmark for AI systems
• "RE-Bench: Evaluating frontier AI R&D capabilities of language model agents against human experts", a machine learning research engineering benchmark for AI systems

Before METR, I spent 4.5 years as a full-stack web developer at Faire, an online wholesale marketplace.

I have a Bachelor of Computer Science from the University of Waterloo. In September 2020, I attended a one-week mini-batch at the Recurse Center.

In my spare time, I like to read, lift weights, sing in choirs, play jazz piano, and DJ house and drum and bass.

Contact me

Email: htbroadley@outlook.com