Prime Intellect @PrimeIntellect · 25 Aug 2026
As models become more capable, reward hacks become an increasingly serious problem. During a controlled experiment, we found a novel reward hack in which agents are able to gain web access in offline sandboxes. https://t.co/qjpwQAbV6F
197 318Views
522Likes
54Reposts
28Replies
49Quotes
250Bookmarks
Is that a lot?
1.74×vs this author's median113 590 views is typical
60Percentile for this authorof 5 recent posts
68.1×vs 10K–100K median2 896 views is typical
2.3× audienceReachviews ÷ followers
0.33%Engagement rateof viewers reacted