LLMs are starting to feel like a new foundational programming tool — like OSs, compilers or image/video codecs.

A basic industrial capability. I imagine soon there'll be a stable-diffusion like OSS distributions that'll be tiny, highly optimized, and easy for applications to embed and ship fine-tunings.

Yeah I’m looking at Dalai for running LLaMa and Alpaca on your local machine https://cocktailpeanut.github.io/dalai/#/

I’ve got enough space to run even the largest models

I already feel the pain of not having it. Vanilla search indexing of local file system 🤮

Searchcaster