Seldon

All of my original ideas are generally 3-4 sentences long. I then read around them, deliberately think on them and write longer substack length articles. I represent individual ideas as nodes, bunch of similar ideas create a graph, and there are different types of edges to represent how exactly two ideas connect. When i add ideas from completely different topics, they cluster around in different graph. My 16 diary system is graph of graphs here.

Traversing through graph to generate articles

This is an hypothesis i am testing. I run different types of traversal algorithms on graph to certain depth and ask LLM to take that sub-graph and write an article using all nodes. All nodes are connected using edges so a second order ideas should naturally form(is the hypothesis). I want to see if different types of traversals generate clearly different types of articles. It is a running experiment, I will write on it after i am done running few iterations.

Today, i gather my individual thoughts and write one article a week. With this system running, i will have traversal articles already written. I will now have not just individual ideas but also cluster of related ideas. I will be able to have this second order thinking done for me to consume, which might in turn enable me to have even longer ideas(which i call third order ideas).

research rabbit

A partially similar tool.