Part 9 — Interview Prep

The rest of this book is for building. This part is for the week before an interview, when building is over and the useful activity is being able to produce the small pieces cold and talk about them.

The mini-projects in Parts 1 through 6 are the full versions — a real ReAct agent, a real tool framework, a real eval harness. Nobody writes those on a whiteboard in forty minutes. What you actually get asked for is the twenty-line core: the loop, the decorator, the retry, the gate. This part is those cores, each with the theory that survives a follow-up and the words to say it in.

Thirty-eight entries across six practical sections and one theory-only section, each in the same shape: what it is in a sentence, the theory in a short paragraph, the smallest correct implementation, the part people get wrong, and a “Saying it out loud” block written the way a person actually talks.

Use it by covering the code and writing it from memory, then covering the spoken block and saying your own version before you read the one given. The gap between what you produced and what is written is the thing to work on — and it is almost always the same two: the failure mode you did not name, and the tradeoff you did not state.

Ends with a thirty-minute night-before routine and the ten questions you should be able to answer cold.

The second chapter, Questions asked in real loops, is the other half of the same idea. It collects questions that people report being asked in real interviews, written down as they were asked, and answers them in full: what the question is really testing, the answer, the follow-up that comes next, and a short spoken version. Questions are grouped by the loop they came from and keep the asker’s own numbering, because a loop has a shape and the shape tells you what the team runs in production. Two loops are in there now — eighteen questions — and it is a running collection, so it grows as more get reported.