Blog

Hermann Ebbinghaus was meticulous to the point of obsession. In the 1880s, he spent years memorising lists of meaningless syllables — "DAX", "BUP", "LOC" — then testing himself at precise intervals to measure exactly how fast he forgot them.

The result was the forgetting curve: a mathematical model showing that memory decays exponentially without reinforcement. Within 20 minutes of learning something, you've forgotten roughly 40% of it. Within a week, up to 90% is gone.

What Ebbinghaus also discovered — and what most people miss — is that the curve can be reset. Each time you actively recall something, the curve flattens. The next time you forget, it happens more slowly. Repeat this enough times and the knowledge becomes practically permanent.

This is the scientific foundation for spaced repetition. The question isn't just "how do I learn this?" — it's "how do I make sure I don't lose it?"

RetainIQ uses the Ebbinghaus model directly. Every skill gets a live retention score calculated as R = e^(-t/S), where t is days since last practice and S is your personal stability score for that skill. When the score drops below a threshold, you get an exercise. When you complete it well, the curve resets and your stability improves.

The goal isn't to keep you studying forever. The goal is to compress maintenance to the minimum time required — then get out of your way.

Ask anyone who spent two years learning Spanish in college and hasn't used it since. They'll tell you they've "forgotten most of it." They haven't, actually — the knowledge is still there, compressed and dormant. What they've lost is retrieval fluency.

The $6 billion language learning market is almost entirely focused on acquisition: learning new things. Duolingo teaches you new words. Babbel teaches you new grammar patterns. Rosetta Stone teaches you new vocabulary in context.

None of them are designed for the person who already knows 2,000 Spanish words and wants to make sure they still know them in five years.

This gap extends far beyond languages. A developer who learned React two years ago and has been doing backend work since. A musician who practiced scales daily for a decade and then stopped. A finance analyst who passed CFA Level 2 and never needs to prove it again — but wants the knowledge to stay sharp.

Maintenance is a fundamentally different problem from acquisition. The content is already there. The goal isn't to add more — it's to preserve what exists with the minimum possible effort.

RetainIQ was designed around this. You don't learn on RetainIQ. You maintain. The AI generates exercises from skills you already have, calibrated to exactly what you're at risk of losing right now. Three minutes. Targeted. Done.

Piotr Wozniak built SM-2 in 1987. It powered SuperMemo, which he used to memorise tens of thousands of facts. Decades later, the same algorithm is at the heart of Anki, the most popular flashcard app in the world.

The core idea: every time you review a piece of knowledge, the system adjusts how long it waits before showing it to you again. Answer well, and the interval grows longer. Answer poorly, and the item resets to a short interval.

SM-2 tracks two values per item: the interval (days until next review) and the ease factor (a multiplier that grows or shrinks based on your performance). The formula for the next interval looks like this:

Next interval = Current interval × Ease factor

The ease factor starts at 2.5 and adjusts based on how well you scored. A score of 5 (perfect) increases ease slightly. A score of 0–2 resets the interval entirely and reduces ease.

RetainIQ uses a hybrid approach. We use SM-2 to schedule when exercises appear — combined with the Ebbinghaus forgetting curve to calculate your actual retention score in real time between exercises. This means your retention score reflects true decay, not just "days since last review."

The practical result: if you practice a skill consistently, intervals between exercises grow from 1 day → 3 days → 7 days → 21 days and beyond. The system learns your personal memory stability for each skill and adjusts automatically.

The single biggest predictor of whether someone maintains a habit is friction. The harder something is to start, the less likely it is to get done. This is especially true for maintenance tasks — things that feel optional until they suddenly aren't.

A 30-minute study session has too much friction for daily maintenance. You have to block out time, get into the right headspace, sit down properly. Most days it doesn't happen.

Three minutes is different. Three minutes happens between meetings. Three minutes happens while waiting for coffee to brew. Three minutes happens because it's easier to just do it than to think about whether to do it.

But there's a floor too. One-minute exercises aren't rigorous enough to genuinely test recall. They're too easy to game — to pattern-match without actually retrieving the underlying knowledge.

The research on desirable difficulties (Robert Bjork, UCLA) suggests that retrieval practice needs to create a small amount of cognitive effort to be effective. Too easy and the memory trace isn't strengthened. Too hard and you disengage.

Three minutes, in practice, is long enough for a genuine open-ended recall task — translating a paragraph, writing a function from memory, explaining a concept without looking it up. It's short enough that the barrier to starting is essentially zero.

This is why every RetainIQ exercise is capped at roughly three minutes. Not because it's an arbitrary round number — because it sits at the exact intersection of rigorous enough to work and short enough to actually happen.