Glossarie bets embedded translations in eBooks can teach languages; 363 HN points

Glossarie is a language-learning app that takes an English eBook and progressively swaps words and short phrases into French, Italian or Spanish at a chosen difficulty, with tap-to-see translations, grammar explanations and practice built around the text.

Words learned in context stick, so replacing text in books people already read can teach a language — if scalable NLP/LLM accuracy improves without rules explosions.

Founder jonathanb88 spent over two years building Glossarie with his partner, a language teacher, before the March 2024 Show HN. His premise: vocabulary is easier to remember in context and with regular repetition, and reading needs no carved-out time — other apps demand a habit of exercises or games, whereas lots of people already read books. The free app targeted French, Italian and Spanish from English.

The launch thread mixed enthusiasm with sharp pushback. Commenters praised the concept and filed feature requests, while critics asked how word-by-word replacement avoids teaching wrong word order — one screenshot rendered French 'mort' after the noun in English order — noted similar ideas in Readlang, Toucan and LingQ, and hit platform gaps: iOS 17+ only, Play Store links failing outside supported regions, and no visible pricing. The founder answered that the real tradeoff is optimizing repetition of vocabulary against quality of grammar examples, called the app 'very much at a proof of concept stage', and positioned it as a free MVP to gather feedback.

Concrete traction: 363 points and 155 comments on the 2024-03-24 Show HN, with commenters testing the app during the thread.

The founder named a falsifiable bet: in-context translation works only if scalable NLP/LLM accuracy gains beat diminishing returns and rule explosions.

The design attacks habit friction — users read books they already wanted to read instead of adding a daily app ritual — a different wedge from flashcard and exercise apps.

Cost structure is unusual: content is precomputed once, so LLMs add capability (explanations, accuracy checks) without per-user API costs at reading time.

Build on habits users already have (reading) instead of demanding a new one, and state the scaling assumption — here, accuracy must improve via scalable NLP, not rules — so the bet stays testable.

As of 2024-03-25, when the Show HN discussion wound down, Glossarie was a free live MVP: an app for learning French, Italian and Spanish from English eBooks, with EPUB uploads and a public-domain library, restricted to regions where the founder had cleared book copyright. The founder collected a long feature list (dynamic difficulty, content-linked practice, browser support, more languages) and promised no charges for current features. The source material records no pricing, funding, revenue or later state, so nothing beyond the thread is claimed.