Tweet by michellechen

May 21, 2026

reverse engineering openai’s goblin problem: we took open models and trained them with RL to talk about goblins an experiment by @willccbb and me, trained on @PrimeIntellect. here's an interactive blog of how RL works and how we achieved goblin mode https://t.co/3KjkdXr9RL

Author
michellechen
Date
May 21, 2026