About Vocady — who makes it and what it is built on

Vocady is a vocabulary app for iOS and Android. You handwrite a word during class; the app reads it on the device, fills in the rest of the entry, and brings the word back later as a question you have to answer from memory.

What the app does

Handwriting is recognised on the device by ML Kit, usually within one to two seconds, across more than 300 languages. Once a language model has been downloaded, recognition runs fully offline.

The Claude API then fills the entry: the dictionary form of the word, IPA, part of speech, meaning, a grammar note, two example sentences with translations, and a CEFR level. Japanese and Chinese entries carry ruby readings alongside the characters.

A photo can register up to 50 words in one pass.

The original stroke coordinates and their timing are stored, so tapping the handwriting in a word detail view replays it in the order and at the speed it was written.

Data is written locally first and then synced across devices.

Daily Recap asks at most ten questions a day. One word produces one question, and no word is asked twice in the same session. All five question types are retrieval tasks: choosing the meaning, filling a blank, choosing the word from its meaning, assembling a sentence, and writing the word out by hand.

When there are not enough plausible distractors for a question, it drops from four options to three, then to two, and finally to handwriting. The question gets harder, not easier.

The scheduling weight is: a base of 10.0, 15.0 for a word never studied, plus (5 − mastery) × 3.0, plus (wrong ÷ attempts) × 5.0, plus min(days since last seen, 10).

The interface is available in nine languages.

Why the studies are published here

Every design choice above is an argument about how memory works, and an argument can be wrong. So the studies behind those choices are listed with their numbers, including the numbers that cut against what a vocabulary app would like to be true.

Kobayashi (2005)

Source and statistics

A meta-analysis of 57 studies and 131 independent samples. Taking notes versus not taking notes gave a weighted mean of d = 0.22. The act of writing something down, on its own, is a small effect.

Kobayashi (2005), Contemporary Educational Psychology, 30, 242–262

Norouzian, Jin & Webb (2025)

Source and statistics

A reanalysis of 27 studies and 55 effect sizes. Conventional note-taking came out at 0.275 (p = 0.299, not significant), while keeping a vocabulary notebook came out at 1.852 (p = 0.003), the largest value in the set.

Norouzian, Jin & Webb (2025), The Modern Language Journal, 109(1), 171–193

Flanigan et al. (2024)

Source and statistics

Across 49 effect sizes, notes that were reviewed gave g = 0.421 and notes that were not gave g = 0.208, Q(2) = 7.94, p = .019.

Flanigan et al. (2024), Educational Psychology Review, 36:78

Dunlosky et al. (2013)

Source and statistics

Of ten study techniques, only two were rated high utility: practice testing and distributed practice. Rereading was rated low.

Dunlosky et al. (2013), Psychological Science in the Public Interest, 14(1), 4–58

Rowland (2014)

Source and statistics

Across 159 effect sizes, testing with feedback gave g = 0.73 and testing without feedback gave g = 0.39.

Rowland (2014), Psychological Bulletin, 140(6), 1432–1463

Cepeda et al. (2006)

Source and statistics

Across 317 experiments with 14,811 participants, spaced practice recalled 47.3% against 36.7% for massed practice.

Cepeda et al. (2006), Psychological Bulletin, 132(3), 354–380

These figures are reported one at a time, next to the study that produced them. Nothing on this site adds or multiplies effect sizes from different studies into a single claim about how much better you will do.

What handwriting is, and is not, evidence for

It would be convenient for an app built around handwriting to claim that handwriting beats typing. The meta-analyses do not support that claim.

Source and statistics

Voyer et al. (2022) report g = −0.008, and Urry et al. (2021) report g = 0.04, which passed a test of equivalence to zero.

Voyer et al. (2022) · Urry et al. (2021)

Wiley & Rapp (2021), Psychological Science, 32(7), 1086–1103, tested something narrower with 42 adults: learning the shapes of unfamiliar letters. Writing scored 91.0, typing 78.5, and looking at the letters 64.5. On spelling words that had not been trained, writing scored 76.3 against 62.3 for typing. What that study measured is the acquisition of letter forms, not the understanding of a lecture.

So handwriting is in this app for two stated reasons: it is what a student is already doing in a notebook during class, so capture costs nothing extra, and there is evidence for it in learning the shapes of an unfamiliar script. Not more than that.

Who makes it

Vocady is built by Yeonseok Kim, an independent developer. There is no research group behind it, and none of the studies cited on this site were conducted by me.

What I did do is read them, check each number against the original paper or the publisher’s abstract, and leave out any figure that could not be traced to a published source. Where a study contradicts what the app does, the contradiction is stated rather than dropped.

If a number on this site misreads its source, please write to dev.tastekim@gmail.com and it will be corrected.