I'm fascinated by the promise of LLMs. A handful of my friends are GPT power users, and I've been using generative models for ~5 years at this point, so I feel like I have a decent sense of what they're capable of and what they're not good at.

What I don't have a good sense of, is how to use them in my own workflow. The reason for this is that I fundamentally don't trust the centralized providers of LLM technology not to create a shadow profile of me, and then to use that for something nefarious.

So, last week I worked on an enterprise implementation of a document analysis tool at a hackathon with my day job. It was awesome, and I understand more about how feasible this is on my own machine now. And I intend to implement such a tool with my journals and notes, locally, with the intent of building myself a collection of agents and helpers.

Anybody else doing this?