SuperMemo: Piotr Woźniak and the first app that really works (1990)

SuperMemo is a learning method and a family of computer programs created by the Polish researcher Piotr Woźniak, first released for MS-DOS in 1987. It is widely regarded as the first spaced repetition software — the first program that did not merely display exercises but calculated, for every single item a learner studies, the optimal moment for the next review, based on a model of human forgetting. Its early algorithm, known as SM-2, was published openly and became the de facto standard of an entire industry: it still runs, in adapted forms, inside Anki, Mnemosyne and countless other flashcard applications. If PLATO (1960) gave computer-assisted language learning its interactivity, SuperMemo gave it a memory model — the first practical answer to the forgetting curve described by Hermann Ebbinghaus in 1885.

Woźniak and the experiments of 1985

In the early 1980s Woźniak was a student in Poznań with a problem familiar to every language learner: the English vocabulary he collected so carefully kept evaporating. From 1982 he maintained paper databases of English word pairs, and — dissatisfied with revising them at random — he began to measure his own memory. In the summer of 1985 he ran the experiment that started it all: working with pages of roughly forty word pairs, he tested how long the gap before a review could grow before recall broke down. The finding was simple but consequential: after each successful review, the next interval can be made considerably longer without increasing the risk of forgetting.

The result was a paper-and-pencil procedure known retrospectively as Algorithm SM-0: review new material after about one day, then seven, then sixteen, then thirty-five, with intervals roughly doubling from there. Woźniak studied with this paper system for two years. In 1987 he got access to his first PC — an Amstrad PC 1512 — and in December of that year wrote the first computer version, SuperMemo 1.0 for DOS. The move to software changed something fundamental: instead of scheduling whole pages of vocabulary, the program could now schedule every item individually, according to how well that particular item was actually remembered.

SM-2: an easiness factor for every item

The algorithm carried by the first DOS versions of SuperMemo — described in detail in Woźniak’s 1990 master’s thesis Optimization of Learning and later published in full on the SuperMemo website — became known as Algorithm SM-2, and it is disarmingly simple. Every item gets its own “easiness factor” (E-Factor), a number that starts at 2.5. After each review the learner grades their own recall on a scale from 0 to 5, and the grade nudges the easiness factor up or down. The first interval is one day, the second six days; after that, each new interval is the previous one multiplied by the item’s easiness factor. An item that is failed goes back to the beginning of the cycle and returns quickly.

In human terms: intervals grow roughly exponentially, easy material drifts far apart while difficult material keeps coming back, and every flashcard ends up with a personal review schedule derived from the learner’s own performance. That is the whole trick — and it works, because it approximates the shape of the forgetting curve: each well-timed review flattens the curve a little more, so ever-longer gaps become safe. SM-2 does not make memorization effortless; it makes the effort land at the moments when it does the most good.

An open algorithm that became the industry standard

Woźniak published the SM-2 formula rather than guarding it, and that decision shaped everything that followed. When open-source flashcard software appeared in the 2000s, its authors did not have to invent a scheduler — they implemented SM-2 from the published description. Anki’s classic scheduler is a modified SM-2; Mnemosyne uses another SM-2 variant; hundreds of commercial and academic learning tools have done the same. By most accounts SM-2 is the most widely deployed review-scheduling algorithm in history — a rare case of a piece of 1980s student research becoming invisible infrastructure for global self-education.

Mainstream fame arrived in 2008, when Gary Wolf profiled Woźniak in Wired in the article “Want to Remember Everything You’ll Ever Learn? Surrender to This Algorithm” — a portrait of a researcher who had structured his entire life, including his sleep, around the optimization of memory.

The research did not stop at SM-2. SuperMemo’s own versions moved on: SM-4 (1989) was the first algorithm that adapted itself to the user, SuperMemo 6 (1991) fitted actual forgetting curves to the learner’s data, and SM-17 (2016) and SM-18 (2019) rebuilt scheduling on a two-component model of memory, distinguishing a memory’s stability from its momentary retrievability. Outside SuperMemo, the open-source FSRS algorithm (2022) applied machine-learning parameter fitting to the same problem and was added to Anki in 2023 as an alternative scheduler. Yet the simple SM-2 remains the workhorse of the field.

SuperMemo today

In 1991 Woźniak and his university friend Krzysztof Biedalak founded SuperMemo World in Poznań to commercialize the method. The software was a finalist of the “Software for Europe” competition at CeBIT in 1992, and the company went on to publish language courses on CD-ROM and, later, online. Today the brand spans the supermemo.com learning platform and the desktop program SuperMemo 18, which runs Algorithm SM-18. Since 2000 the desktop edition has also offered incremental reading — a technique, invented by Woźniak, for reading thousands of articles in parallel by extracting fragments and feeding them into the spaced repetition queue.

Woźniak himself largely withdrew from business into research. He has published extensively on memory, sleep and learning — including his long-standing practice of free-running sleep — and maintains his writings openly at supermemo.guru. One historical footnote matters for fairness: spaced repetition as a research finding is much older than SuperMemo, and Woźniak’s contribution was not the discovery of the spacing effect, but the first working computational machinery that put it into everyday practice.

What this means for language learning

Vocabulary is the part of language learning where forgetting hurts most: thousands of small, similar items, each fading on its own schedule. This is exactly the problem SuperMemo solved. An SM-2-style scheduler guarantees that no word is silently lost — every item comes back precisely when it is about to slip away — while keeping the daily review load bounded, because well-known material recedes into month- or year-long intervals. Virtually every serious vocabulary tool built since — from Anki to modern spaced repetition systems in language apps, Taalhammer among them — schedules its reviews with an algorithm descended from SM-2. What SuperMemo could not provide was scale and convenience for the mass market; that arrived when smartphones put a spaced repetition scheduler in every pocket, an expansion symbolized by Duolingo (2011).

Frequently asked questions

Was SuperMemo really the first spaced repetition app?

Spaced repetition itself is older: Ebbinghaus described the forgetting curve in 1885, Cecil Alec Mace ran the first practical experiments in 1932, and Sebastian Leitner popularized a paper flashcard box in the 1970s. SuperMemo (1987) was the first software implementation — the first program that computed an individual review schedule for every item from the learner’s actual performance, which is why histories of spaced repetition software conventionally begin with it.

Is the SM-2 algorithm still used today?

Yes. Anki’s classic scheduler, Mnemosyne and a large share of flashcard apps still run SM-2 or close variants of it, almost four decades after its creation. Newer algorithms exist — SuperMemo’s own SM-17/SM-18 and the open-source FSRS, adopted by Anki as an option in 2023 — but SM-2’s simplicity, published specification and decades of practical validation have kept it the de facto standard.

Does SuperMemo still exist?

Yes. SuperMemo World, founded in Poznań in 1991, operates the supermemo.com learning platform, and the desktop program SuperMemo 18 continues to ship Woźniak’s latest algorithm, SM-18, together with incremental reading. Woźniak himself continues to publish research on memory, learning and sleep.

Sources:

1. Piotr A. Woźniak, Optimization of Learning, Master’s thesis, Poznań University of Technology, 1990: https://super-memory.com/english/ol.htm
2. Algorithm SM-2 — SuperMemo: https://super-memory.com/english/ol/sm2.htm
3. General history of SuperMemo — super-memory.com: https://super-memory.com/english/history.htm
4. Gary Wolf, “Want to Remember Everything You’ll Ever Learn? Surrender to This Algorithm”, Wired, 2008: https://www.wired.com/2008/04/ff-wozniak/
5. SuperMemo — Wikipedia: https://en.wikipedia.org/wiki/SuperMemo
6. “What spaced repetition algorithm does Anki use?” — Anki FAQs: https://faqs.ankiweb.net/what-spaced-repetition-algorithm
7. “The true history of spaced repetition” — supermemo.com: https://www.supermemo.com/en/blog/the-true-history-of-spaced-repetition