Show HN: Phrasing – learn every language, to any level

Original Article ↗ Hacker News Discussion ↗

A new language-learning app that builds study materials from parallel texts, TV shows and other media is drawing interest for its use of GPT-powered tooling and modern spaced-repetition algorithms, but also scrutiny over quality and clarity. Commenters highlight serious issues with example translations (even in major languages), the ambitious “learn every language” promise, and a landing page and playground that make it hard to understand what the product actually does. Many see potential in features like pre-learning key vocabulary from shows, but argue the service must focus on reliable, human-vetted content, clearer communication of language support, and better onboarding before it can be trusted for serious study.

Product concept & target users

  • App turns parallel texts (e.g., essays, TV shows, subtitles) into language-learning material: aligned sentences, vocabulary extraction, audio, SRS.
  • Currently oriented toward advanced learners/polyglots who already understand concepts like “shadowing” and parallel texts; clearer explanations for beginners are requested.
  • Vision is “bring your own content” rather than supplying full courses.

Landing page, UX, and communication issues

  • Many users say they “don’t understand what this is” from the playground or main page.
  • Confusion over the TV‑show feature: unclear whether it streams content, dubs, or just augments subtitles.
  • Demo interactions (clickable tags, search box) are not self-explanatory; some mistook demo search for a live content search.
  • Pages are reported as laggy/slow on mobile and desktop; playground loads slowly; some layout/input jank noted.

Translation quality & machine translation concerns

  • Multiple examples (Turkish, Spanish, French, Italian, Hungarian, Japanese) are reported as ungrammatical, literal, misaligned, or semantically wrong.
  • Several users conclude they “wouldn’t trust” the tool based on these samples.
  • Maintainers say demo translations come from “official” or external sources; some missing parts were auto‑translated.
  • Strong skepticism about any reliance on GPT/OpenAI for translation, especially for non‑English targets and low‑resource languages; Duolingo’s quality drop cited.
  • Concern that automatic mining of “parallel texts” on the web will ingest poor, machine‑translated material.

Language coverage and “every language” claim

  • Tagline “learn every language” is challenged: many languages (e.g., Albanian, Bengali, Haitian Creole, Esperanto, various regional languages) are missing.
  • Users want an explicit list of supported languages and a per‑language support status (e.g., alpha/beta/full).
  • Some push back on breadth‑first strategy; suggest doing a few language pairs very well with curated, human‑checked content.

Learning methodology & SRS discussion

  • Debate over what’s “hardest” in language learning: choosing what to learn vs. finding engaging, level‑appropriate input vs. staying consistent.
  • Interest in the idea of pre‑learning key vocabulary from a specific show/episode to boost comprehension.
  • Spaced repetition: app claims a more “humane” FSRS‑based system that handles breaks better than classic SM‑2/Anki defaults.
  • Thread includes side discussion on memory curves, Anki’s algorithms, and comprehensible input vs. phrase memorization.

Technical issues and performance

  • Many users cannot sign up due to “Email rate limit exceeded” / “Load failed”; tied to external auth provider downtime and/or rate limits.
  • Email input fields are sluggish or non‑editable on some mobile browsers.
  • Some users see temporary content loading failures and a confusing env.js debug artifact.

Related tools and positive reactions

  • Several commenters share similar projects (readers, podcast tools, SRS apps) and express enthusiasm for the core idea, especially:
    • Using real media, especially intermediate‑level input.
    • Integration with SRS and personal content.
  • Despite criticism, UI/UX design and the general concept receive praise, alongside encouragement to keep iterating.