Retain med school lecture volume: the math nobody runs before MS1.

Run the arithmetic before you pick a method

A representative US pre-clinical week:

• 4 to 6 hours of lecture per day, typically split into 2 to 4 slide decks.
• 10 to 20 decks per week, each one 50 to 120 slides.
• 80 to 160 decks per 8-week block, all examinable, mostly cumulative.

If a single deck of cards takes you 60 to 120 minutes to author by hand, the bookkeeping is brutal. Two decks a day is 2 to 4 hours of authoring before you have done one review. Four decks a day is a whole second school day. Across a block that is 80 to 320 hours of card-writing, which is structurally impossible to spend in addition to attending lecture and sleeping.

That is the part most retention advice never names. Active recall does not fail in MS1. The authoring tax in front of active recall fails, and what looks like a discipline problem in week 5 is a queueing problem from week 1.

Recognition feels like retention. It is not.

Watching a lecture at 2x and rewatching it is the most common survival reflex when the queue gets ahead of you, and it is the worst use of the same hour. The testing-effect literature, going back to Roediger and Karpicke 2006 and replicated many times since, puts retrieval practice at roughly two to three times the retention of equal-time rereading or rewatching, measured a week to a month out.

An hour of rewatching a 60-minute lecture finishes one deck and trains recognition. An hour of answering 60 to 100 questions on that lecture finishes the deck, trains recall, and produces a review queue that survives the next block. Same hour, very different week-three outcome.

The workflow that survives a 16-week pre-clinical course

Four steps, repeated per deck, repeated per day. The trick is the per-deck cost: if step 1 is cheap, the whole loop runs. If step 1 is expensive, the loop dies on a busy Thursday.

Hand-rolled Anki versus a per-deck generator, line by line

Anki itself is not the problem. The DIY authoring pipeline in front of it is. Here is what changes when the per-deck cost goes from an hour to a minute, and what stays the same.

The anchor: questions come from your professor's actual deck

This is the part you can verify on one upload. Drop a single 90 to 120 slide lecture deck into Studyly. In about 60 seconds it returns roughly 200 questions across four formats, multiple-choice, free-response, case-style, and image-occlusion, all generated against that exact deck. Not a generic web question bank, not a shared community deck someone else made for a different professor. The wording, emphasis, and edge cases mirror what your block test will pull from, because they came from the same source the test will.

On revisit Studyly auto-rephrases the prompt and rotates the distractors, which removes the lazy first-three-words pattern-match that quietly kills retention in week 2. Each deck also grows its own tree as you drill it. By exam-day morning the screen reads as a forest, one tree per deck, and a stunted tree is the only honest signal that a lecture is under-rehearsed.

A wrongly-keyed question is worse than no question

Any tool that writes questions for you, ChatGPT included, will produce confident, plausible answers with no quality check. For retention at volume that is a tax: a wrong key gets repped just as often as a right one, and a misnamed pathway carries to the block test. Question quality matters more in MS1 than in most other study contexts, because the same flashcard gets seen 6 to 10 times across a block.

Read that as the company's own measurement on a consistent rubric, not an independent audit. Factual correctness is scored first on purpose: for a pre-clinical retention workflow it is the criterion that decides whether the next 8 weeks of review queue is an asset or a slow-burning liability.