The story
Four years ago, the goal was to learn Mandarin. Not the phrasebook version -- the real thing. Read a newspaper. Follow a conversation. Understand what was being said in the Chinese dramas playing in the background.
It started where everyone starts. Duolingo. It was fine for a week. The green owl kept the app opening, the multiple-choice questions were easy enough to feel productive, and the streak counter gave a daily sense of accomplishment. Then came the attempt to read a menu in Chinese -- and not a single character could be identified. The progress bar had been measuring engagement with the app, not engagement with the language.
Next came Anki, the spaced repetition tool that serious language learners swear by. Anki is powerful. It is also, essentially, a blank engine. You have to build the car yourself. More time went into formatting flashcards -- should there be audio? Pinyin on the front or back? Both recognition and production directions? -- than into studying Chinese. The tool was excellent at scheduling reviews. It offered nothing for content, curriculum, listening, tones, or the dozen other skills that make up actual Chinese ability.
HelloChinese, Hack Chinese, textbook courses came after that. Each solved one piece. None solved the problem that kept coming up every single day: studying vocabulary in one place and trying to read Chinese in another, with an enormous gap between those two experiences. Nothing connected studying to encountering. The act of looking up a word during reading and the act of drilling vocabulary were separate activities, happening in separate tools, with no feedback loop between them.
That gap became the product. A text editor was opened and building began -- not a startup, not a side project, a tool that was actually needed. Python. SQLite. A spaced repetition scheduler adapted from FSRS, the open-source algorithm that represents the current state of memory science research. Drill types: not just flashcards, but tone pair discrimination, cloze deletion, audio-to-character matching, sentence construction, register-aware practice. Forty-four types in total, each targeting a specific cognitive skill, each grounded in research on how memory forms and retrieval strengthens.
And then the core feature, the one that solved the original problem: the cleanup loop. Read Chinese in the graded reader. Look up a word you do not know. That word automatically enters your drill queue, scheduled across multiple drill types. The thing you encountered during reading becomes the thing you practice during drilling. The loop closes. Reading and practice are no longer separate activities.
Today, Aelu is a complete adaptive drilling and reading system, used daily by its creators and a growing community of learners. When something does not work, it gets noticed. When something improves, that gets noticed too. The distance between the builders and the users is zero.
