Spaced repetition: how to stop forgetting what you read.
Everything below is settled science that almost nobody manages to apply. This page is what the research says, what the intervals should actually be, and the specific reason most people give up in the first month.
What spaced repetition is
Spaced repetition is reviewing something at growing intervals instead of all at once — tomorrow, then in three days, then in a week, then in a month — with each review timed to land just before you would have forgotten it.
Two things are doing the work there, and they are usually confused with each other. The spacing is the gap between reviews. The retrieval is what you do at each one: trying to produce the answer from memory before you check it. Spacing without retrieval is just rereading on a schedule, and rereading is close to worthless.
Why forgetting is not a discipline problem
In 1885 Hermann Ebbinghaus spent a year memorising nonsense syllables and measuring how fast he lost them. The shape he found has not changed since: retention drops steeply within the first day, then flattens out low. Roughly 58% survives twenty minutes, about a third survives a day, and around a quarter is what remains after a week.
It is tempting to dismiss this as Victorian data about meaningless syllables. Murre and Dros repeated the experiment in 2015 and got the same curve. Meaningful material decays more slowly than nonsense, but it decays in the same shape — which is why the talk you found genuinely brilliant is gone by Thursday.
The important consequence is not that you forget. It is thatyou cannot tell what you have forgotten. Nothing signals the gap. You will keep believing you know the thing until the moment you need it.
The two findings that make it work
The testing effect. In Roediger and Karpicke's 2006 experiment, one group read a passage four times and another read it once and then tested themselves on it three times. Immediately afterwards the rereaders felt better prepared and scored slightly higher. A week later the testers remembered 61% and the rereaders 40%. The act of retrieving is not a way of checking learning; it is the learning.
The spacing effect. Cepeda and colleagues reviewed hundreds of studies in 2006 and found the same pattern throughout: identical study time spread across days beats the same time in one sitting, and the advantage grows the longer you need to retain the material. Cramming works for Friday. It does not work for the job.
How long should the gaps be?
The rule of thumb is that the optimal gap is roughly 10–20% of the time you need the memory to last. To remember something for a month, review it after three to five days. To remember it for a year, review it after a month or two.
In practice a schedule that starts at one day and multiplies by about 2.5 each time you recall correctly — 1 day, 3 days, a week, two weeks, a month, three months — is close enough to optimal that the difference stops mattering. What matters far more is what happens when you get one wrong: that card goes back to the start. An interval you failed is evidence the gap was too long.
This is also why a fixed weekly review session does not work. Every item is on a different clock. On any given day, some things are due and most are not, and reviewing the ones that are not due is wasted effort — the spacing effect only pays out near the edge of forgetting.
Why almost everyone quits
None of the above is controversial, and tools that implement it have existed for twenty years. Anki is free, open source, and scheduled by a well-tested algorithm. So why does a method with this much evidence behind it have so few long-term users?
Because the method sends you a bill before it pays out. Every hour of watching or reading turns into an evening of writing cards: splitting ideas into single facts, phrasing each as a question with exactly one answer, keeping the wording consistent enough that you are testing the memory and not the phrasing. Good cards are a genuine skill, and bad cards make the whole system worse — you end up drilling ambiguous questions for months.
So the failure is not a discipline failure either. People quit spaced repetition for the same reason they quit most systems: the cost lands today and the benefit lands in three months.
Doing it without making cards
The obvious fix is to remove the evening. That is what badNerd is: you save a video, an article or a note, and the AI pulls out the ideas worth keeping, writes the questions, and puts them on the schedule described above. Every card keeps a link to the exact place it came from — the timecode, the paragraph — so checking one is a second's work, and you can edit any of them.
The reviewing still has to be yours. Retrieval cannot be outsourced; that is the part that does the work. But five minutes of it a day is a habit people keep, and an evening of card-writing is not.
Get started freeFree while in beta · nothing to install
Sources
- Ebbinghaus, Memory (1885) — the original savings-method curve, replicated by Murre & Dros (2015).
- Roediger & Karpicke (2006) — repeated retrieval beats repeated study at one week.
- Cepeda et al. (2006) — meta-analysis of distributed practice across 254 studies.