[kingkongjaffa]

If I wanted to go from 0 to up and running I would read https://simonwillison.net/2023/Dec/31/ai-in-2023/

But honestly just playing with the openly available tools (openAI, Anthropic, Ollama running locally, lang chain) you can quickly figure out the basics at least for inference and some RAG techniques.

With that knowledge you should be able to evaluate features being built by companies and have a decent idea of if they are likely feasible using the current LLM’s.

I personally have been building RAG Assisted text generation tools, I’m yet to see anything else like agentic LLM’s that can use tools be useful yet.

The next step for me is understanding better RAG techniques to build more queryable embeddings, and figuring out good methods for structured outputs. E.g. likely you dont want just plain text output you want JSON or something, reliably.