AI writes a lot of code now. What's left for us, and what interviews increasingly test, is working with code we didn't write.
That's three skills:
- Debugging someone else's code from a bug report.
- Adding a feature without breaking what already works.
- Reviewing a pull request and deciding whether it can merge.
Most practice sites still ask you to write an algorithm from scratch. So I built DevDrill, an open-source app that runs on your own machine.
How it works
You get a small, realistic project in a VS Code-style editor in your browser, in JavaScript or TypeScript.
- Debug: a bug report, a few files and failing tests. Find the cause and fix it. When you submit, hidden tests check the edge cases, so patching only the symptom usually fails.
- Feature: working code and a ticket with an exact API contract. Build it without breaking the existing tests.
- Review: a pull request with its diff. Comment on the lines that matter, approve or request changes, then compare with the answer key.
There are 17 challenges, including an Express pack. Nothing leaves your machine and there's no account. AI is optional: connect your own provider for hints or for judging your review comments.
Try it
git clone https://github.com/KennyCandra/devdrill.git
cd devdrill
npm install
npm run dev
Then open http://localhost:5173/.
It's MIT licensed, and new challenges are easy to add. I'd love to hear which challenges feel too easy, unfair or unclear.
Top comments (0)