+19% slowdown on average was observed in the first METR RCT study published in July 2025: "Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity"
The spirit of this market is to compare both studies apples-to-apples. Hopefully METR will officially tell us what the second study's equivalent percentage figure is, that can be directly compared with the first study's +19%. If not, this market will resolve N/A.
The bounds are left inclusive and right exclusive except for "+15% to +20%", in which right is also inclusive.
If no second study is published before 1st August 2026 this market will resolve N/A.
Update 2025-11-19 (PST) (AI summary of creator comment): If METR reports results for both a new cohort and a subset similar to the original cohort, the market will resolve based on the subset comparison (if METR is comfortable making that comparison). Otherwise, the market will resolve N/A if only the full new cohort results are reported.
1,000People are also trading
@CharlesFoster If the subset is similar enough that METR is comfortable doing a comparision, market should resolve based on that, otherwise N/A if only the uplift of the full new cohort is reported.
Ugh, I'm starting to understand how they get a slowdown. Several times recently I've asked ChatGPT to read the docs for something and tell me why my code isn't working, it hallucinates something nonsensical, often making up parts of my code that don't exist. Then I read the docs myself and the answer is right there.