METR’s Joel Becker on exponential Time Horizon Evals, Threat Models, and the Limits of AI Productivity

589 snips

Feb 27, 2026

Joel Becker, researcher leading METR (model evaluation and threat research), explains the origin and limits of the Time Horizon chart. He walks through how tasks were chosen and biases in benchmarks. The conversation covers developer productivity studies, milestone-like model jumps, threat modeling for discontinuous takeoff, and trade-offs in building evaluation scaffolds versus real-world deployment.

Ask episode

AI Snips

Chapters

Transcript

Episode notes

Opus 4.5 Prompted Engineers To Embrace Agentic Coding

Joel recounts Opus 4.5 as a large benchmark jump that many engineers began using for agentic coding.
He observed talented engineers shift from avoiding AI for code to relying heavily on it after that release.

Redo Productivity Trials Only With Updated Study Design

Re-run productivity RCTs carefully because workflows and selection effects changed as models improved; prior designs may no longer be valid.
Joel warns of developer selection bias and increased concurrency in modern workflows disrupting previous randomization.

Apparent Developer Speedups Often Overvalue Low‑Value Work

Perceived speedups often overstate value because extra tasks enabled by AI are lower value than previously postponed work.
Joel notes concurrent work, selection, and user optimism inflate self-reported productivity gains.

Get the Snipd Podcast app to discover more snips from this episode

Get the app

This is a free preview of a paid episode. To hear more, visit www.latent.space

AIE Europe CFP and AIE World’s Fair paper submissions for CAIS peer review are due TODAY - do not delay! Last call ever.

We’re excited to welcome METR for their first LS Pod, hopefully the first of many:

METR are keepers of currently the single most infamous chart in AI:

But every Latent Space reader should be sophisticated enough to know that the details matter and that hype and hyperbole go hand in hand in AI social media, because the millions of impressions that got, by people who don’t understand or care about the nuances, disclaimers, and error bars, far outreaches the 69k views on the corrections by the people who actually made the chart:

There’s a lot of nuance both in making benchmarks (as we discovered with OpenAI on our SWE-Bench Verified podcast) and in extrapolating results from them, especially where exponentials and sigmoids are concerned. METR’s Long Horizons work itself has known biases that the authors have responsibly disclosed, but go far too underappreciated in the pursuit of doomer chart porn.

If you’re interested in a short, sharable TED talk version of this pod, over at AIE CODE we were blessed to feature Joel twice, as a stage talk and with a longer form small workshop with Q&A:

We also make sure cover some of METR’s lesser known work on Threat Evaluation but also Developer Productivity, where 2x friend of the pod and now Zyphra founder Quentin Anthony was the ONLY productive participant!

Finally, if you’re the sort to read these show notes to the end, then you definitely deserve some pictures of Joel shredding the guitar at Love Band Karaoke which we mention at the end:

Full Video Pod

Timestamps

00:00 What METR Means00:39 Podcast Intro With Joel01:39 ME vs TR03:33 Time Horizon Origin Story04:56 Picking Tasks And Biases09:13 Time Horizon Misconceptions11:37 Opus 4.5 And Trendlines14:27 Productivity Studies And Explosions29:50 Compute Slows Progress30:47 Algorithms Need Compute32:45 Industry Spend and Data34:57 Clusters and Shipping Timelines36:44 Prediction Markets for Models38:10 Manifold Alpha Story43:04 Beyond Benchmarks Evals51:39 METR Roadmap and Farewell

Transcript