Sign in to keep reading
“The Same Hour, Failed” is free with an account. Your progress and bookmarks follow you across devices.
Sign in to continueNEW: ML Mock & Coaching with a 20+ year experience coach. Save $20 on sessions for Premium.ML Mock & Coaching now available
NEW: ML Mock & Coaching with a 20+ year experience coach. Save $20 on sessions for Premium.ML Mock & Coaching now available
An equally capable candidate, the identical brief, forty minutes, and a no-hire. The four decisions that separate the two runs — none of which required knowing more.
“The Same Hour, Failed” is free with an account. Your progress and bookmarks follow you across devices.
Sign in to continueWhich One Is You?
Could you answer these today? Tick what you already know.
Same brief. Same file. Same forty minutes. A candidate who knows Python at least as well, and who could find this bug in thirty seconds reading it on paper with nothing at stake.
No hire. Here is how.
0:00 — Selects the file, pastes it into the chat panel, types why is this broken?
0:02 — The model replies confidently that the refill logic looks standard, and suggests the real risk is thread safety. It is not wrong that thread safety is a risk. It is wrong that it is this bug. The candidate starts adding a lock.
0:06 — Locking is not fixing anything. Re-prompts: no, requests are getting through that shouldn't. Back comes a restructured class using a sliding window instead of a token bucket. The candidate accepts it — it is longer and looks more sophisticated, and accepting it feels like momentum.
0:14 — Still wrong. Pastes the failure back in. Third version arrives. The file on screen no longer resembles what the interviewer handed over, and the candidate could not now explain what most of it does.
0:21 — The interviewer asks: "So what was the original bug?"
The candidate says something about the timing being off. It is not wrong, exactly. It is not an answer either. This is the moment the round is decided, though it takes another nineteen minutes to end.
0:28 — Asks for per-user support. Sixty lines arrive. Accepts them. Asks for tests. They pass on the first run, which feels like relief rather than a warning.
0:36 — "What happens to memory as users accumulate?" The candidate scrolls through their own file looking for the answer.
0:40 — Time. The code arguably works. The interviewer has almost nothing to write down.