lmao the prompt injection is funny. my experience is more on ML (prev recommender systems + current lm fine-tuning/evals). happy to chat more if you foresee such a need!