Hugo Larochelle @hugo_larochelle · 17 Feb 2026
Ever since @METR_Evals' fascinating work on LLM agent time horizons, I've wanted to see other attempts to draw conclusions from the same data. In a separate approach by Fengyuan and Jay, we too infer an exponential, but with shorter horizons for recent models (~2h vs METR's ~5h)
6 386Views
21Likes
4Reposts
3Replies
0Quotes
8Bookmarks
Is that a lot?
0.43×vs this author's median15 017 views is typical
0Percentile for this authorof 5 recent posts
1.54×vs 100K–1M median4 138 views is typical
5.11%Reachviews ÷ followers
0.44%Engagement rateof viewers reacted