AI Coding Interview Practice — AI skill for Claude Code

AI Coding Interview Practice

*Ten timed problems for the AI-assisted coding interview, run in your own editor with your own agent.*

Amazon, Google, and a growing list of companies are moving to an interview format where an AI assistant is allowed for the whole session. You get a small codebase with bugs, a solver to implement, and timed tests that force you to optimize. The interviewer is not grading whether you can write the algorithm from memory. They are grading how you drive the assistant.

This repo puts that setup in front of you locally, with whatever agent you already use: Claude Code, Cursor, Codex, Copilot. One problem per folder, an answer key per problem, and a prompt that turns your agent into the interviewer.

The four phases

flowchart LR
    A["1. Comprehension
read a small codebase
and map how it fits together"]
    B["2. Bugs
fill in the ???? values,
follow the failures to two bugs"]
    C["3. Implementation
make the solver correct,
brute force is fine"]
    D["4. Optimization
uncomment the timed tests
and climb the ladder"]
    A --> B --> C --> D

Every problem runs the same way the real interview does, in about 50 working minutes. The domain class ships with two planted bugs. Some tests are commented out with `????` where the expected value should be: fill those in, run them, and follow the failures. Then make the correctness tests pass. Then uncomment the timed tests, which brute force will not survive, and find the better algorithm, and then the one after that, because the last timed test usually breaks the first optimization too.

Quick start

1. Clone the repo.

   git clone https://github.com/azizu06/ai-coding-interview-practice
   cd ai-coding-interview-practice

2. Pick a problem from the table below and read its README, for example [`problems/01_word_container/README.md`](problems/01_word_container/README.md).

3. Start a 50 minute timer, open your coding agent