Finally filled out my forecasts. This was excellently designed!
Loved having the context laid out since I'm too lazy to read the papers. TL;DR I think everything saturates except FrontierMath, which merely dectuples.
Talk about it at Minifest tomorrow!
Conversation
I highly recommend filling out this survey about 2025 AI capabilities before it closes Dec 31, for epistemic glory and edification! theaidigest.org/2025-ai-foreca
If you're coming to Minifest tomorrow, I'm hosting a workshop at 11 AM to work on 2025 agent benchmark forecasts. At the end I'll also ask about prob of extreme AI outcomes (full cognitive automation, full cognitive+physical automation, Dyson swarm, AI takeover) by 2040.
Seems like we basically agree everything gets saturated (I had OSWorld lower because humans were weirdly low and I thought maybe many of the tasks were broken).
I started a newsletter with zero writing skills and no clue about newsletters.
But I took the plunge anyway.
Here's how the first official full month went:
What does it all mean?
Who knows!
But I'm loving the growth! 
Want to be boring with us?
Show more
“With long timeline people like these, who needs short timeline people”
I never felt like a "long timelines person" tbh, since 2019 my attitude has been "Holy crap this could be soon & we should prep." Left tails matter more than medians! And IME a lot of people who have "shorter" timelines than me forecast much weaker endpoints.