I’m experimenting with an OpenAI-based assistant powered by my own writing. It is wild. See this transcript of me testing it with a couple questions. It does hallucinate sometimes, but as a powerful search or just for fun, this could be something.

@pratik Yes, essentially what I’m doing is using the OpenAI “assistants” API to feed it an export of my whole blog. I’ll blog about it when it’s further along. There are other tools that can do this but I’m forgetting the names right now.

the Assistants API is pretty much my favorite way of working with OpenAI’s models. they’ve done a great job of making RAG simple.

this is interesting. I have some local LLM’s on my laptop I can run with Ollama. I might have to try giving it a few years of my blog as well and see what happens.
