Show HN: Duolingo-style exercises but with real-world content like the news

Overall reception

• Many commenters find the core idea—Duolingo-style exercises based on real videos—very appealing and more engaging than synthetic sentences.
• Several say they would use or pay for it, especially as a supplement to Duolingo; others find it currently too rough or difficult but see strong potential.

UX and interaction

• Drag‑and‑drop is widely criticized, especially on mobile: hard to target, words reorder after dragging, and early auto‑grading feels punishing. Many request simple click‑to‑fill and a “submit” button.
• Autoplaying, looping videos divide opinion: useful for repeated listening, but others find the default loop annoying and confusing (controls not clearly tied to the video).
• “Number of gaps” is confusing jargon; “blanks” or hiding this upfront is suggested.
• Requests include: keyboard shortcuts (space to pause, arrows to seek), clearer advancement for manual input (e.g., visible “Next” button), stable word-bank layout, better error highlighting that shows what the user actually chose.

Content selection and difficulty

• Difficulty is inconsistent: some beginners find clips impossibly fast; others find certain “news” clips too slow or classroom‑like, especially in Japanese.
• Strong demand for:
  • Explicit difficulty levels (including speech speed).
  • Beginner modes and slow‑news sources.
  • Topic filtering (science, AI, gossip, etc.).
• Concerns about news as a source: videos can be depressing or polarized; some worry about agenda‑pushing and want more neutral or educational channels.
• YouTube subtitles and on‑screen text can sometimes give away answers, weakening the exercise.

Language coverage and quality issues

• Several per‑language quirks:
  • Japanese: need furigana, kana‑only options, better word segmentation (e.g., not splitting conjugations into two blanks), and timing fixes where the target word is cut off.
  • Spanish and Finnish: transcription/normalization bugs cause correct answers to be marked wrong.
  • Portuguese: currently Brazilian; users request clearer labeling and European variant.
  • Requests for many new languages (Mandarin/simplified, Greek, Swedish, Irish, Swahili, etc.).
• Commenters suggest labeling languages as alpha/beta until quality stabilizes; smaller‑language ASR is notably weaker.

Learning features and pedagogy

• Many want in‑app translations:
  • Sentence translation after answering.
  • Per‑word meaning on click.
  • Ability to save words/phrases and build vocab decks.
• Some propose adaptive difficulty based on known vocabulary, spaced repetition, and idiom‑aware translations.
• A few see potential for crowdsourcing better transcripts and using this data to improve models.

Technical and platform considerations

• The app relies on YouTube; some users hit corporate firewalls, certificate issues, or “prove you’re not a bot” blocks.
• LLM‑based content filtering (“non‑war, non‑politics”) can still surface low‑quality or ideological clips; users call for more explicit editorial control and disclosure of AI filtering.

Monetization and product direction

• Current one‑time payment is seen as refreshing versus subscriptions, but some doubt it will cover ongoing transcription costs.
• Multiple users urge “non‑enshittified” monetization and suggest clearly showcasing the backlog of Pro exercises to justify paying.

Flags and representation

• A long side discussion critiques the use of national flags to represent languages (e.g., Union Jack for English, Brazilian vs European Portuguese).
• Many argue flags conflate nation and language, which can be misleading or offensive; alternatives like ISO language codes or neutral labels are suggested.