If I had to guess, Q* would involve solving problems quantum-style, by breaking them down along dimensions (not step-by-step!!), then solving the subtasks in pa

If I had to guess, Q* would involve solving problems quantum-style, by breaking them down along dimensions (not step-by-step!!), then solving the subtasks in parallel and recombining. Basically, most of the LLM intelligence is at the word-to-word selection level, so friction needs to be created there instead of at the task or question level.