← The conversationsLatent Space Podcast
METR’s Joel Becker on exponential Time Horizon Evals, Threat Models, and the Limits of AI Productivity
Listen to this conversation56 min
Read transcriptThanks for reading
Stay a little ahead of the curve.
Fresh ideas, original reporting, and deep conversations in your inbox.
Subscribe to Latent SpaceGo a little deeper
The wikiSourced profiles and separately reviewed explanations connected to this post.
Person
Joel Becker
AI evaluation researcher whose work at METR examines the gap between benchmark performance and useful software development.
Organization
METR
An AI evaluation nonprofit measuring autonomous capabilities, real-world productivity and risks from frontier developers’ internal use of agents.
Person
Quentin Anthony
AI systems engineer whose work spans distributed model training, practical training guidance, and hardware-aware model design.
Grounded writeup
What METR’s Time Horizon Measures—and What It Leaves Out
Human task duration makes AI capability easier to interpret. Productivity, autonomous research and catastrophic risk depend on more than the work a benchmark captures.