2024 LLM Journal Arcs

This journal and my professional life are getting a reboot (LLMs!). It’s a space for reflection, for marking time, and for organizing my daily journey. January has been designated as my immersion month, and I am already amidst the chaos of various half-understood topics and the creative zeal of individuals pushing the boundaries. The frontier is in constant motion. This journaling endeavor is tasked with outlining the threads I am following, providing some continuity in a shifting landscape.

The first arc is chatting with my documents. There’s real value in making my personal data accessible to LLMs, and yet no set answer regarding how to expose that data for my own personal documents—much of the data isn’t accessible to the Googles/OpenAIs of the world, e.g., physical photos, notebooks, recordings. I’ll be laying out my work and product thoughts for the next two weeks.

I’ve been trying tools in this space since August, starting with raw ChromaDB, getting frustrated by LangChain, excited by LlamaIndex, entranced by MemGPT, and tantalized by Mamba. What I’m trying to get at is I’ve got some ideas for posts:

• How are documents chunked?
• LlamaHub and LlamaIndex
• Competitive and collaborative analysis
• MemGPT and how it’s breaking my budget
• My own indexing setup, esp. with org.email, emacs frontend a la GTD-shell

Come February I’ll have a better idea how I want to operate in this space and these journal entries’ll be on to the next arc.