The 11-Interview Debrief — The Full Tape, the Exact Scripts, and the Feedback I Could Not Publish on Medium
The Medium version was the log. This is the autopsy. Round-by-round transcripts, the exact 90-second openings I now run, the rejection emails verbatim, and the three things I refuse to do in interview
This is the paid edition. The Medium version covered what happened. This one covers exactly what I said, exactly what they said, and what I would do differently with another rep at each one. I am keeping it behind the paywall because some of the scripts are still working for me and I do not want them on the open internet for ATS-trained models to pattern-match against.
If you are preparing for senior backend interviews right now and you read the Medium piece thinking “fine, but what did you actually say in interview 5 that worked,” this is that.
Why a paid version exists
I got asked the same three questions in the Medium comments and DMs:
What did your scoping opening actually sound like?
Can I see one of the rejection emails verbatim?
What did interview 10 look like end to end?
The honest answer to all three is “I am not putting that on a public post that 8,000 people read.”
The reason is not modesty. It is that the scripts in this post are still in active rotation in my own interviews and a fair number of yours. If I publish the exact 90-second opening in the open, three things happen. ATS-adjacent screening tools train on it. Interviewers see it five times in a week and start scoring against it. And it stops being a senior signal and becomes a tell.
So the deal is simple. You pay $5 for the month, you get the full debrief. I keep the part that is still load-bearing in my own job hunt behind a wall where only the people who actually want to use it will see it.
Section 1 — Interview 2 verbatim
The Medium piece said I drew boxes for forty minutes and got rejected for “senior-level scoping.” Here is what actually happened, with the timestamps from my notes.
The prompt (paraphrased to protect the company): “Design a system that notifies users about events relevant to them, at scale.”
Minute 0 to minute 3. I asked one clarification question: “Push, email, in-app, or all three?” Interviewer: “Up to you.” I took this as license to design for all three. This was the first mistake. “Up to you” was not license. It was the first test, and the test was whether I would scope.
Minute 3 to minute 8. I drew the producer side. Event bus, fanout, routing rules. I named Kafka. I named the consumer pattern. I felt good. The interviewer was quiet. I now know quiet is bad.
Minute 8 to minute 20. I drew the channel-specific delivery layer. Push service, email service, in-app service. I drew a database for user preferences. I drew a deduplication layer. I drew a retry queue with exponential backoff. I knew this part cold. I kept drawing.
Minute 20 to minute 30. I drew the read side. How users see their notifications in the app, how the unread count works, how the read receipt flows back. The interviewer asked one question. “What scale?” I said “Let’s say a hundred million users.” The interviewer wrote something down. I now know what they wrote down. They wrote down that I had picked the scale arbitrarily, twenty-two minutes into the interview, because I had not picked it at the beginning.
Minute 30 to minute 40. I tried to discuss tradeoffs. The interviewer let me. The follow-ups were short and one-word. “Why.” “What else.” “Anything else.” I did not realize at the time that these were rescue prompts and I was failing every one of them.
The rejection email, verbatim, minus identifying detail:
