How spaced repetition works in Krendo.

Spaced repetition is the unglamorous reason any vocabulary tool works. The idea is older than computing: instead of cramming a word ten times in a row, you space the encounters apart and increase the gap as the word sticks. This page explains how Krendo implements that in plain language, what triggers a status change, and why we made a few specific choices that differ from a vanilla SRS like Anki.

Three statuses

Every saved word in Krendo has one of three statuses: new, learning, or stronger. (There are two more — ignored and archived— for words you want out of the review queue without deleting; we’ll ignore them here.)

• Newmeans you’ve saved the word but haven’t reviewed it yet. It hasn’t earned a place in your memory.
• Learningmeans you’ve gotten the word right at least once but not consistently. It’s in the window where consistent encounters compound, and that’s where the spaced-review queue lives.
• Strongermeans you’ve gotten it right multiple times in a row, with non-trivial gaps between reviews. It’s reliably yours. Krendo still surfaces it occasionally — long-interval reactivation — but not in the daily queue.

What moves a word between statuses

A correct answer (in flashcard mode, you hit “I knew it”; in self-rate, you mark “easy” or “OK”; in multiple choice, you pick the right option) promotes the word toward stronger. A wrong answer demotes it —back to learning if it was stronger, or it stays in learning if it was already there.

Crucially, “stronger” isn’t earned by one right answer. Krendo looks at the recent answer history and the time gap between encounters. Three correct answers in a row, with at least a few days between them, is the rough shape. The point is that you can’t bluff your way to stronger by drilling the same word ten times in five minutes. The intervals are doing the work.

Why intervals matter more than streaks

The most common misunderstanding of spaced repetition is to treat it as a streak game — review every day, build a long chain, feel rewarded. The science says something more specific: the value of a review is highest when you’re about to forget the word. If you review too often, you’re refreshing what was still fresh. If you review too rarely, you’re re-learning what you already lost.

Krendo’s review queue tries to surface words at roughly the right moment. Words you just learned reappear soon (within a day). Words you’ve gotten right twice reappear in a few days. Words at stronger status appear weeks apart for long-interval reactivation. When you miss one, the interval resets back to the start of the curve.

Three review modes, one underlying state

You can review the same queue three ways:

• Flashcard shows the prompt, you click to reveal the answer, you tell Krendo whether you knew it. Closest to traditional flashcards.
• Multiple choiceshows the prompt with four options — one correct, three distractors sampled from your library’s same-target-language words. The distractors avoid words with the same tags to keep the test honest.
• Self-rateshows the prompt and the answer together, and asks you to rate how well you knew it (easy / OK / hard / again). Good for vocabulary where the “binary right or wrong” framing is too coarse.

All three modes write to the same status state. You can switch between them mid-week and progress is consistent.

What Krendo does differently from vanilla SRS

A vanilla SRS treats each card independently and uses pure interval math (FSRS, SM-2, etc.) to decide what to surface. Krendo does some of that, but adds two opinions that come from how people actually learn vocabulary from real reading:

• Topical clustering avoidance. When picking the next newword for a session, Krendo skips one whose tags overlap with the last three already shown. Adjacent new words sharing a tag — e.g. three business terms in a row — cause semantic-clustering interference and slow acquisition. The rule applies only to new words; learning and stronger words can cluster freely.
• Sentence context is part of the card. Because every saved word carries the sentence it came from, the front of the card you see in review isn’t just the word — it’s the word in the sentence in the language it was used. The recall cue is the same cue you’ll get when you re-encounter the word in the wild.

Practical recommendations

From watching how the tool gets used:

• Five minutes of review a day beats forty minutes twice a week. The intervals only do work if they fire.
• Don’t over-save. If a word is too easy, ignore the Save button — you don’t need a card for it. Krendo isn’t scoring you on how many words you save.
• If a word keeps failing, look at the sentence context. Often the context is too thin (a one-word headline, a list item) and the cue isn’t strong enough. Save the same word from a richer sentence later — Krendo will append the new context, and review will use the most recent one.
• Use Anki export when a word graduates from learning to stronger and you want it in your long-term retention deck. The two systems are complementary; Krendo doesn’t try to be the be-all-end-all retention engine.