I think the old report you're referencing is this [1] from July 2025, but I can't find a new report. This [2] links to a new dataset at the bottom (that maybe shows improvements?) but it seems like they chose not to write it up because of perceived flaws in their study. Is there a more relevant report I'm missing?

[1]: https://metr.org/blog/2025-07-10-early-2025-ai-experienced-o...

[2]: https://metr.org/blog/2026-02-24-uplift-update/#wider-adopti...