“A report by Andy Matuszak demonstrated that LLMs struggle to write effective spaced repetition prompts, a task that requires projecting a user's future mental state.”