Dan Hendrycks @hendrycks · 21 Aug 2026
More findings that interpretability tools are fragile or worse than simple baselines.
9 651Views
81Likes
1Reposts
8Replies
0Quotes
43Bookmarks
Is that a lot?
1.19×vs this author's median8 082 views is typical
60Percentile for this authorof 5 recent posts
3.33×vs 10K–100K median2 896 views is typical
21.48%Reachviews ÷ followers
0.93%Engagement rateof viewers reacted