Brainstorming with LLMs

Currently attempting to vibe-code a harness that sets Claude up as a brainstorming partner with heavy human oversight. Based around a user-defined corpus of existing thought (a Karpathy Wiki for now, but I’m not sure that’s the best approach), I’m experimenting with morphological analysis, synectics, and some other methods to combine related and unrelated ideas in interesting ways…and in greater quantity that I could realistically do on my own…with multiple model steering points along the way.

Screencap showing direction setting via morphological analysis (with made up data in this case).