Anki + vignette MCQs as a daily habit: the atomic-card conflict and how to actually resolve it

Treat each vignette as one atomic card, where 'atomic' means one underlying concept-pin rather than one sentence. A 130-word NBME-shaped stem can still be atomic at the concept layer if the answer key traces to exactly one fact. Then size your daily reps by minutes, not card count: a vignette card takes 25 to 40 seconds to read and rate, vs about 5 seconds for a classic basic card. Twenty new vignette cards a day is roughly the same time budget as 100 new basic cards a day. Finally, attach a per-deck reward that does not reset on miss, so a skipped day pauses progress instead of erasing it.

It violates the surface-level reading of the rule. The deeper rule is one card, one thing to learn. A 130-word vignette that ends in 'which segment of the nephron is responsible' is atomic at the concept layer: the entire wrapping resolves to one fact about the thick ascending limb. The version that genuinely breaks atomicity is a stem that asks two things ('most likely diagnosis and best next step'), because the card now tests two facts on one rating. Studyly's case-style generator emits one-fact-per-stem by default; a multi-part vignette would be split into two cards with two separate topic-pins.

Cap by minutes first, then convert backward to a card count. If you have 15 minutes for daily reps, that is roughly 25 to 30 vignette cards including the reading time and a brief explanation read. New cards turn into reviews on a schedule, so 20 new vignettes a day becomes a steady-state load of 80 to 120 cards a day at FSRS 0.90 retention after about six weeks. Most med students who burn out on Anki do it by setting new cards by count without converting to minutes; for vignettes that arithmetic ends in a 90-minute daily queue in week eight.

As MCQ cards. The export carries a custom note type, studyly_mcq, with four option fields, an Explanation field, and a studyly_card_id field that lets a re-import match cards by ID instead of by stem text. The front template renders the stem and the four options as a list; the back template highlights the correct one and shows the source span. Anki schedules each note as one card under whatever scheduler you have set (FSRS is the default in current Anki releases). The Anki documentation on note types is at docs.ankiweb.net/getting-started.html if you want the long form.

In Studyly the stem is regenerated on revisit against the same topic-pin: different demographic, different presenting symptom that maps to the same underlying mechanism, rotated distractor pool. In Anki the .apkg fields are static once imported, so the stem stays put unless you re-export from Studyly and re-import. The supported pattern is to use Anki for cards where the rote wording is the point (drug names, anatomy labels, vocabulary) and stay in the Studyly app for the cards where stem rotation matters most. The two are complementary, not exclusive.

The failure mode is asymmetric feedback. In stock Anki, showing up empties the cards-due count to zero, then it refills tomorrow with nothing carried over visually. Skipping a day grows the queue, so the only number you ever see move is the punishment one. In Studyly each deck grows a tree across about two weeks: day 1 is a bare trunk at 0.18 height, day 3 has 2 leaves, day 7 has 4 leaves, day 14 has 7 leaves at 0.92 height. The growth schedule lives in the open source at src/components/TreeGrowth.tsx in the marketing site repo. A missed day pauses the tree at its current height; it does not reset to a bare trunk, which is the specific failure mode that makes streaks backfire.

You can. The cost is about 4 to 8 minutes per card if you also write four parallel distractors, an explanation, and a source citation. Across a 90-slide lecture deck that is somewhere between 6 and 13 hours of authoring before you drill the first card. The .apkg export path converts the same deck in about 60 seconds and arrives with the same fields populated. The trade-off is editorial control: if you want to phrase every stem yourself, hand-authoring is the right call; if you want drillable cards same-day-as-lecture, the export is the right call.

Use UWorld. The reason to also drill vignettes from your own lecture deck is coverage: UWorld is written against the published Step 1 content outline, not against what your professor decided to emphasize this block. Your block exam, your shelf prep, and the half-mark questions on your real Step 1 that come from a topic UWorld treats lightly all reward drilling on your professor's actual slides. Lecture-grounded vignettes close that coverage gap; UWorld closes the writing-quality gap. The Anki habit layer is what carries either set into long-term retention.

No. The habit framing assumes you have weeks of daily reps before the exam. On cram night the scheduler is already as compressed as it gets, and the right question is whether the cards you are drilling test what the exam will test. Two hours on a 200-card vignette deck pinned to your professor's slides usually beats four hours on a 400-card recognition-only basic deck. Studyly's tree mechanic on cram night functions as a focus aid (something growing on screen makes the slog bearable) rather than as the long-term reward it is built to be.