中文 Literacy Speedrun II: Character Cyclotron

For reading Chinese, 90% token coverage wasn't enough, and 99% level is believed to be needed for comprehension.

To avoid constantly getting stuck on low-frequency words, the approach shifted away from learning-while-reading toward forcibly drilling symbols into the head.

The core bottleneck was that on the flashcard site Hack Chinese, information was scattered all over the place.

• The front of the card shows the word, the back only shows pronunciation and meaning
• Etymology, stroke order, and sub-character information are in a separate dictionary window
• Every time you want to check a word, you repeat copy-paste, clicking around, scrolling, and asking an LLM for explanations
• Every step is slow, breaking the learning flow

The solution was to build an extension that injects information directly into the browser so you never have to leave the flashcard window.

Claude Code was tasked with the development to quickly produce a large amount of JavaScript, connecting Unicode character graph and phonetic mapping data to instantly pull up etymology, components, pronunciation, and calligraphy information.

What mattered here was immediacy rather than adding more features.

• Matched the panel's border radius, background color, font, and alignment to the existing UI
• Added keyboard shortcuts like E for etymology, C for calligraphy, T for tones
• Allowed selecting characters by number to reduce navigation cost
• Ran the morphology LLM call and dictionary lookup concurrently to hide wait time

As a result, a single lookup that used to take more than 30 seconds now takes under 1 second, and keystrokes are no longer the bottleneck.

In the end, information started coming in so fast that they say it's reached the point of wondering whether the brain itself needs to be optimized next.