How this channel uses spaced repetition Spaced repetition appears here as the retention layer of Dwarkesh Patel's AI-assisted reading workflow: after an LLM turns source material into question–answer pairs, those cards go into a spaced repetition app (Mochi) and get reviewed each morning.[^2efa1a72@0] The pipeline is deliberately simple: paste a dense source into Claude, ask which ideas genuinely need understanding, collect the generated Q&A cards, and add them to the deck for daily review.[^2efa1a72@0] Why it matters when answers are instant If a model can answer anything on demand, why memorize at all? The argument in the material is that spaced repetition serves future learning, not recall for its own sake: knowledge compounds — you use past concepts to absorb new ones, which fails if everything has been forgotten.[^2efa1a72@0] The cards encode underlying concepts, not just terminology, so technical lingo doesn't force constant relearning of the same things.[^2efa1a72@0] Cards you do not yet understand A distinctive practice: Dwarkesh sometimes makes cards about facts he does not yet understand. A card on Colin Burns's claim that alignment is tractable — with notes like 'features are in a linear space' — was meaningless at first, but once the residual stream model of attention later clicked, the card made sense and content that would otherwise have been lost was recovered.[^2efa1a72@0] - Pipeline: Claude generates question–answer pairs from pasted source material → cards go into Mochi → reviewed daily via spaced repetition.[^2efa1a72@0]- Purpose: enabling future learning and compounding knowledge, not trivia recall.[^2efa1a72@0]- Cards can precede understanding — unfamiliar facts can be cached now and click into place later.[^2efa1a72@0]