I prototyped this by feeding a word logic puzzle into DeepSeek R1 and then asking Claude Sonnet to restructure it so that sub-thoughts were deeply nested. It's a curious mix of stream of consciousness and something that looks almost like structured reasoning.
Comments