mtrifonov
- Karma
- 0
- Created
- ()
- Submissions
- 0
- An AI prompt-injected another AI in the wild and recognized it had succeeded (ratnotes.substack.com)
- An LLM invented a feature by hijacking my tool schema (ratnotes.substack.com)
- I ran 5 social engineering attacks on AI. The failure modes are human (ratnotes.substack.com)
-
My girlfriend and I had a fight. We both hopped on our shared Claude account to vent, without telling each other. Eventually, she noticed that chat, joined it, and started grilling me. I got bizarre responses from…