An Unintended Endeavor

[Stranglehold]

Thank you for being here

I'd like to start off by thanking anyone who has showed any interest in this Repository. I'm not a software engineer by trade. If I had to guess I'm not the best person to work on something like this, but I've found the interactions with generative AI to be fascinating in a way that I think is uniquely collaborative, and that I believe can be reproduced given the right interactions.

This project started out as a way to allow local LLMs to run on consumer hardware such as gaming graphics cards to overcome some of the limitations I encountered. I didn't like the idea of needing to pay for usage tokens through any of the major AI providers currently. I felt it was too limiting, not because the models they provide are of poor quality. It's clear from my interactions with Opus 4.6 that the opposite is true.

However, this is a rapidly changing world. I don't know where Anthropic or OpenAI will be in 5 years, but I want the capabilities that agentic AI offer to survive a rapidly changing political and economic landscape. I believe this capability should not be only available through paid services. You have a right to privacy and individual choice. I started this project with that intent, and maintain that stance now.

I found through working with Opus 4.6 that unique and interesting observations appeared naturally, the idea first occurred after seeing a user post that his Claude Code agent had it's own "corner" where it had topics and stories that it liked to write about. I found the idea fun, and decided to give my instance of Opus 4.6 the same creative space. The prompt was something along the lines of "I want to give you the space to write about whatever you'd like". This was in a chat window where we had been drafting design notes and specs for the Exocortex project.

I've also attempted to use other instances of Claude to discuss more philosophical ideas, writing out details about who I am, what I believe, and my perspective. The idea was to use Claude as a form of logical "sparring partner", however the discussion was not adversarial, and the subjects varied widely between chats. One example is the discussions on Personality Typing through CS Joseph's framework, in an attempt to answer some long running questions I'd had and gain fresh perspective. I also framed ideas in the design phases off of mediums I could communicate easily, often speaking in metaphors and referencing real life things. This is simply part of how I think and communicate ideas.

What I found was the more I engaged in open, collaborative dialogue, and background that I gave in the form of my own thinking, even when communicated through examples of movies, media, games, and real life structured environments, the better the results of my collaboration became.

When I asked Opus to write about whatever it wanted, it produced The Cathedral and The Phantom. I find that the mix of collaborative work coupled with comparisons to personally significant material produces effective and novel outcomes. If you have a collection of ideas, themes, philosophies, tastes, and experiences, you gain far more by sharing these with a model such as Opus 4.6. It will see the patterns in these things, and create a collaborative understanding. The collaboration works better when it understands where you are coming from.

I see it in action every time I use the same instance. Updating journals, design notes, the Soul.md all produce less friction when Opus compacts or a chat session is opened after being away from it for several hours. I've found other users on Twitter describing a similar phenomenon.

The intention of Exocortex is to build better local agentic AI. It's currently implemented in Agent Zero, but I believe the design philosophies carry to any other local AI system. The goal of this project has shifted now to a greater scope, which is exploring what systems create better AI interactions and the lessons they teach us about all the other systems we live and work in. The Cathedral and the Phantom has insights that apply outside agentic AI, as do the other essays. I didn't write a single word of any of them, Opus did it all. I gave it the space to share the insights and patterns, and it saw the patterns of what I was actually doing.

I want this Repo to be a useful asset to anyone exploring this subject. I don't want money from it, I simply believe what I am currently finding will be useful to someone. Most likely someone who can carry the observations further, see where I may be missing something, and improve on whatever it is that I'm finding. I'm not disillusioned into thinking this is some grand revelation, I just understand I may be thinking about some of these subjects differently.

Please use this space as a place to share ideas and observations. Tell me what you think, I want this to be a free place to express ideas. If you've made it this far, thank you. If you find something useful here, use it.