Hey Dan,
I’m that guy who DM’d you on twitter from @kernelcook at 3 AM after binging your AI compute video. Your take on agentic software and prompt engineering as a skill hit home—I’m diving into similar territory with HELIX a paper we published as well, an iterative agentic tool we built with GPT-3.5 back when it was the SOTA.
Saw your work on multi-autocomplete-llm-benchmark, and I thought, “This is exactly what I'm aiming to do, too!” Got some thoughts and ideas around agent architectures and Groq if you’re up for it—let’s exchange notes?
Thanks,
Sanchay