Sublime

was fun to work on this @latentspacepod post w/ @benhylak inspired by "pivot token" literature, one gpt->o1 mental model shift i've made is the role of self-evaluation and -correction as an ESSENTIAL part of planning/reasoning. with o1, you move the LLM-as-judge *INTO THE PROMPT*, so you can let it handle the self eval and replanning. this is t... See more

The best startup AI Engineers I've met are all building their own agents. I know it's a buzzword that's now a bit past the wave of peak hype (thanks @zachtratar), but I can think of no better way to immediately, viscerally, run into all the SOTA problems of Prompt Engineering, RAG, Evals, Tool use, Code generation, Long horizon planning, et al. Y... See more

this is really interesting. seems like she used trial-and-error to converge on the optimal emergent behaviors of the RL for each of these models in the sequence that makes them maximally useful very streamlined and seems minimally lossy