Gokul Rajaram @gokulr · 02 Sep 2026
Cyber RL environments for AI agents If there's one thing that Mythos and the OpenAI / Hugging Face incidents have shown us, it's that AI agents must be treated as compromised before the run begins. Any agent with code execution, credentials, tools, or network access can misuse
17 959Views
42Likes
2Reposts
12Replies
1Quotes
60Bookmarks
Is that a lot?
1.93×vs this author's median9 327 views is typical
67Percentile for this authorof 9 recent posts
0.80×vs 100K–1M median22 338 views is typical
15.01%Reachviews ÷ followers
0.32%Engagement rateof viewers reacted