I’ve stopped just “reading” research papers.
I give them to Codex to unpack the key ideas, trace the math, challenge assumptions, and turn methods into working code. Then we run experiments, debug failures, and iterate.
Paper → prototype → insight, in one loop.
@rasbt Hi @rasbt, the LLM Architecture Gallery keeps getting better — it's an incredibly valuable resource, thank you for maintaining it! 🙌
Out of curiosity, have you built any agentic workflows? Or is the whole process still mostly manual?