wtf is codep?

Codep is a JavaScript tool that turns the waiting time during AI coding sessions into English vocabulary practice. When you use an AI coding agent such as Claude Code, Kiro CLI, Codex, or GAL to work on a project, there is often a pause while the AI thinks or writes code. Codep opens a split terminal pane using tmux and automatically starts a spelling practice session during that pause, then switches focus back to the AI agent once it finishes its task. The tool watches for the AI agent to become busy or idle using hook scripts, then updates its practice interface roughly twice a second. Practice happens letter by letter: type a correct letter and it turns green, type a wrong one and the whole word restarts, in a style similar to the Qwerty Learner project it credits as inspiration. There are two study modes: a quick flashcard style where you recall a word's meaning and rate how well you knew it, and a stricter spelling mode where you type out full words. Word pronunciation comes from the Youdao Dictionary API and is cached locally so it keeps working after the first listen. Codep also plays mechanical keyboard sounds for key presses, correct answers, and mistakes. It ships with built in word lists: about 1700 common programming terms, 2607 words from the CET-4 English exam, and a small custom list users can extend by editing a JSON file. Words are grouped into chapters of 20, and progress saves automatically. Review scheduling uses a spaced repetition system based on the SM-2 method, so words you get right are shown again after longer gaps, while words you struggle with, forget, or peek at using the Tab key return sooner. A mistake notebook automatically collects words you found unfamiliar, fuzzy, misspelled, or peeked at. Multiple difficulty combinations let you hide the word, the definition, or both, for tougher dictation style practice. Installation runs through a single shell script that finds the latest release, clones it, and configures the hooks for your AI agent automatically. It requires Node.js 16 or newer and tmux, and works with the terminal based AI coding tools listed above. Windows users are told to run it inside WSL. The project is released under the MIT license.