I am a second-year nursing student, which means every single week there is more to memorize than the week before, forever, until I either graduate or move to a cabin. Last semester my Sunday ritual was making Anki cards by hand from my lecture notes: copy the term, type the definition, format the card, repeat for ninety minutes. By the time my deck was ready I was too fried to actually review it. I had built a beautiful study tool and then gone to take a nap on top of it.
The cards I made by hand were all the same shape
Here is the embarrassing part. Every card I made was term-to-definition. Front: 'digoxin.' Back: 'cardiac glycoside, increases contractility.' That is fine for a vocabulary quiz and useless for the exam I was actually taking, which asks things like 'a patient with renal impairment is on digoxin — what are you watching for?' One of those tests memory. The other tests whether you understand the thing. I was drilling the wrong one and then acting surprised when my grades didn't move.
So I changed the job. Instead of making cards, I'd make the AI make cards, and I'd spend my energy on the review. I pasted a full chapter on cardiac glycosides into Gemini 2.5 Pro, told it I was at board-exam level, and asked for 18 cards using the Turn Messy Lecture Notes into Spaced-Repetition Flashcards prompt. Three minutes later I had a clean Markdown table.
What a balanced deck actually looks like
The deck it gave me was not eighteen definitions. It was a mix, and the mix is the whole point. There were five definition cards for the vocabulary I genuinely had to know cold. There were six application cards — the 'when would you give this, when would you absolutely not' cards that mirror how the test is written. Three comparison cards forced me to tell digoxin toxicity in an elderly patient apart from a pediatric one. And there was one misconception card that went straight at the thing half my cohort gets wrong: that a normal potassium level means you're safe to dose. You are not.
That misconception card is the one I want to talk about, because it's the reason I started trusting this thing. I would never have written that card by hand. I didn't know it was a common error — that is sort of the definition of a misconception. Gemini handles long context well, so when I fed it the whole chapter it noticed the trap the chapter was warning about and built me a card to rehearse the correct answer until it was reflex. A card that targets the exact mistake you're prone to is worth ten cards that test things you already know cold, and a tired human at the end of a chapter is the last person who can spot which mistake that is.
How I actually run it, step by step
My weekly loop is now embarrassingly simple. I dump the raw lecture notes — typos, half-sentences, the bit where I clearly stopped paying attention and just wrote 'IMPORTANT???' — straight into the prompt. I set the subject, set the level to 'board exam' because that's the bar I'm actually being held to, and ask for 18 cards. It hands back a clean Markdown table with a 'Why it matters' column, which is the column I didn't know I needed. That column is a one-line reason each card earns its place, and reading down it is itself a study aid: it's basically a map of what the chapter wanted me to take away.
- Definition cards lock the vocabulary you need automatic
- Application cards rehearse the exact 'when/why' format the exam uses
- Comparison cards stop you from blurring two similar drugs under pressure
- Misconception cards are worth double — they train you out of the wrong answer everyone else picks
- The STUDY FIRST line at the bottom told me which card was most exam-probable before I'd reviewed anything
Then the cards go into Anki, because the table pastes in almost clean, and Anki's spacing algorithm decides which ones I see and when. That's the other half of the trick the prompt's name promises: spaced repetition. Making good cards is pointless if you cram them all the night before; the whole value is reviewing the hard ones repeatedly over weeks until they stop being hard. The prompt builds the cards. Anki schedules them. I just show up.
The actual result, with the honest caveat
My pharmacology grade went from a B to an A-minus. I want to be careful here, because I don't think a prompt raised my grade. Reviewing better cards more often raised my grade, and I only reviewed more often because I wasn't burning the whole afternoon building the deck. The prompt bought me back the Sunday. I spent it studying. The honest caveat: the first run gave me a couple of cards that were too wordy on the back, so I now tell it '1–2 sentences, self-contained' explicitly, which is baked into the prompt now. I also still read every card before I trust it — twice it pulled a detail that was technically true but not what my professor emphasized, and I cut those. It is a study tool, not an oracle; the moment you stop proofreading the cards is the moment you memorize a confident little error.
The other thing I've learned: the quality of the deck is the quality of what you paste in. Garbage notes make garbage cards, so on the weeks I actually take decent notes the deck is noticeably sharper, which has had the side effect of making me a better note-taker out of pure self-interest. When I'm building a deck for a unit test that's mostly scenario questions, I pair this with my Generate a 10-Question Quiz with a Self-Grading Answer Key prompt so I can dry-run the exam after I've drilled the cards — drill with the flashcards, then prove it stuck with the quiz. Grab the Turn Messy Lecture Notes into Spaced-Repetition Flashcards prompt on Prompt Dock, paste in whatever chapter is ruining your week, and go take the nap after you study instead of before.