What's new in Aelu. Features, improvements, and fixes in reverse chronological order.
Vocabulary coverage expanded from the original 299 HSK 1-3 seed items to over 10,000 items spanning HSK 1-9 (10,978 total). Full curriculum coverage for all nine HSK levels, each item with pinyin, translations, context notes, collocations, and register information.
Full web UI with Flask and WebSocket. The Civic Sanctuary aesthetic -- warm stone, dusty rose, sage -- with automatic dark mode via prefers-color-scheme. Web Audio API for session start and completion sounds. Desktop and mobile responsive.
Record yourself saying a word or phrase and get instant feedback on tone accuracy. Uses macOS TTS for reference audio and grades your recording against the target tone pattern. Available in all sessions where tone practice is scheduled.
Every vocabulary item now includes usage notes, collocations, register information, and common mistakes. 299 context notes across HSK 1-3, so you learn how a word is actually used -- not just what it means in a dictionary.
Sessions now adjust to your available time and energy. A 10-minute evening session focuses differently than a 30-minute weekend session. The system detects your patterns and optimizes drill selection accordingly.
Drill types are now mixed within sessions by default. Instead of blocked practice (all tone drills, then all cloze drills), the system interleaves types to produce stronger retention. Based on research showing interleaved practice outperforms blocked practice for long-term recall.
The system now supports HSK 7-9 vocabulary and drill scheduling. Content seeding is underway. This prepares for the HSK 3.0 standard arriving in July 2026, including the new mandatory speaking component.
Practice with realistic dialogue situations: ordering food, giving directions, making plans, and more. Functional conversation patterns you will actually use, not textbook conversations nobody has.
HSK readiness projections now factor in vocabulary, listening, reading, and tone accuracy independently. Instead of one optimistic date, you get a grounded estimate based on your demonstrated pace across all tracked skills, with confidence intervals.
Daily study streak is now tracked and visible on the dashboard. The momentum indicator shows your consistency trend without guilt mechanics -- no notifications, no shame. Just data about your practice pattern.
A new CLI command that lets you target specific weak areas. Tell the system to focus on tones, listening, or a specific HSK level, and it adjusts your next sessions accordingly. All proposals are executable directly from the diagnostics view.
Vocabulary items now carry automatic tags for register, frequency, and common confusion pairs. The system uses these tags to schedule register-aware and pragmatic drills, so you learn not just what words mean but when and how to use them.
Fixed an issue where the scaffolding system for new items was introducing too many new words per session. Now narrowed to 2 new items at a time, which prevents overwhelm and aligns with research on working memory limits during acquisition.