Articles
YAKKI articles
These articles describe how we test our own decisions: what did not work, what we changed, and what the available evidence can — and cannot — show. Each one is a review of our own mistake, with the numbers it rests on.
Articles
Who decides what a child reads
A ready-made content filter let through all four passages that were plainly unfit for a child: it answers a different question from ours. The article covers what we built instead, why keeping a good text is harder than cutting a bad one, and three findings that matter more than the result.
How we measured text difficulty and fooled ourselves three times
The instrument produced rising, plausible numbers when there was nothing to measure. First it was measuring text length. Then it agreed with itself, because it had been fed the answer. Then it said nothing about grammar, because rising features had been added to falling ones. The article shows how this came to light, and why the method cannot be called our own.
How to simplify Dickens without ruining him
Living children's texts do not supply enough rare grammatical constructions, and we forbade ourselves to compose such texts from nothing: a composed text is plausible and dead. The article shows how we take a real author's text down one step, six traps along the way, and why an absolute scale does not travel from someone else's corpus to ours.
A companion to the textbook In Development · demo
Only a demonstration exists; this is not part of the product yet. It will be an add-on for A1 and A2 for a small monthly top-up in coins; we are filling the textbook shelf. A child learns from one specific textbook, while digital tools offer topics in general — "food", "past tense", "animals". The article shows how YAKKI builds a map of a unit, matches it against its games, and is meant to give practice on the very words printed in that unit. We do not reproduce the textbook and we do not store it; we take nothing that copyright protects.