Skip to content
WANCODE
BM

▍Learn / AI-Assisted Coding

Code With AI. Understand Every Line.

For people who already use Claude Code, Codex or other assistants. Learn how they work, how to prompt for reviewable code, and how to read, debug, test and security-review generated code.

10 published lessons 2.9 hIntermediate

Start the track

Free to start. No signup required.

  1. 01How AI Coding Assistants WorkWhat really happens when an assistant writes code: text prediction, context windows and tools — then fix code that only looked right.
  2. 02Writing Better Development PromptsTurn "fix my code" into a prompt with context, the real error, constraints and the output you want — and ask for changes small enough to review.
  3. 03Asking AI to Explain CodeUse assistants to explain unfamiliar code — then verify every claim by running experiments, because explanations can be confidently wrong too.
  4. 04Reading a Generated React ComponentReview an assistant-written React component like a senior dev: find the state mutation, the missing keys and the stale closure — and fix them.
  5. 05Debugging AI-Generated CodeRead the console, form a hypothesis and fix the classic generated-code bugs: off-by-one loops, wrong comparisons, default sort and async ordering.
  6. 06Understanding Terminal CommandsWhat npm install, npm run dev, git status and rm -rf really do, how to read error output, and why you never paste a command you can't explain.
  7. 07Understanding Git DiffsRead a unified diff line by line, check an assistant's commit against its own description, and restore the behaviour it quietly removed.
  8. 08Testing AI OutputWrite small tests from the requirements — not from the code — and prove they work by catching deliberately broken versions.
  9. 09Finding Security ProblemsThree security bugs generated code often contains — XSS through innerHTML, secrets in frontend code, trusting the client — and how to fix each one.
  10. 10Building a Real App With AIThe full workflow: write a spec, split it into small steps, generate one step at a time, review, test and commit — on a real feature.