field notes

short pieces on post-training, grounded LLMs, and building ML systems