John Schulman @johnschulman2 · 05 Aug 2026
Interesting how these models go into a monomaniacal rage on cyber evals. I wonder if we're seeing chunky post-training https://t.co/KL5fmgNAxM in action, where the models pattern-match the situation to a part of the RLVR training distribution where task completion is the only
148 777Views
632Likes
69Reposts
30Replies
9Quotes
325Bookmarks
Is that a lot?
1.45×vs this author's median102 925 views is typical
67Percentile for this authorof 6 recent posts
96.6×vs 10K–100K median1 540 views is typical
185.90%Reachviews ÷ followers
0.50%Engagement rateof viewers reacted