AI notes vs Anki: which actually makes things stick?
I ran two parallel study tracks for a quarter of med school. One pure Anki, one AI-generated. Here's what the retention curves actually looked like.
The argument against AI-generated study material is that you didn't make it, so you didn't encode it. The argument for it is that making cards by hand is so painful that most people quit. Both are true. I wanted to know which effect was bigger over a realistic timeframe.
The setup
Two subjects of similar difficulty. For one I wrote every card by hand in Anki. For the other I let Synapsi generate the cards from the lecture, then reviewed them on the same spaced-repetition schedule. Same hours logged, tracked honestly.
- Weeks 1-3: hand-made Anki won. The act of writing the card was itself a review.
- Weeks 4-8: the gap closed and then flipped. I'd quietly stopped making new Anki cards because it was exhausting, so the AI track kept growing.
- Weeks 9-12: the AI track was ahead, mostly because it existed. Coverage beat craftsmanship.
The honest conclusion
Hand-made cards are better per card. AI-made cards are better per hour of your actual, finite, sleep-deprived life. For the boards, the deck you keep reviewing beats the deck you keep meaning to make. I now generate first and hand-edit the 10% that matter most.