What We Actually Ask in Interviews
The three interview questions we actually ask, the principles behind our three-hour take-home task, and the patterns in why we turn candidates down.
The Math on Last Month’s Hiring
Last month, April 2026, we read 34 applications, met 8 people, and decided to work with 1. A single interview runs 90 minutes, and since Jamie and I both sit in, that is 3 hours of company time. Eight interviews come to 24 hours, three full days of work. More months than not, we pay this cost and hire no one. So we decided to simply publish what we ask in those 90 minutes, and why. The risk of hearing rehearsed answers is already priced in. These questions are the kind that a prepared answer does not survive for long.
The Three Questions We Actually Ask
First. “Walk us through one problem you held onto longer than any other, starting from the moment you first noticed it.” Most people start with the solution. What we listen for is how long they stay inside the problem before the solution shows up. People who restate the problem in their own words stay there longer. More than once, we have been surprised to find that the shinier the title, the shorter that stretch.
Second. “What did you give up along the way?” Anyone who has seen a problem through has thrown something away. They cut the scope, or slipped the schedule, or swallowed their pride. This spring, one candidate with a dazzling resume answered, “Nothing, really. It all worked out.” Another candidate, with three years at an academy front desk, spent more than 40 minutes on the story of tracking a single payment error for three weeks and ending up rewriting the refund guide. We will not say which of them we decided to work with.
Third. “After it was over, what did you leave behind for the next person?” Someone who solved a problem all the way through leaves traces of the finish. A document, a handover, or at the very least a wrap-up message to a colleague. When the answer to this one comes back empty, we take it to mean the problem was never solved. It just went by.
The Take-Home Task and the Three-Hour Rule
Ninety minutes of interviewing shows us how someone talks, and not much more. That is why we run a take-home task. Our first attempt, early this year, was a failure. We handed out a problem we had actually lived through, nearly unedited, and it turned out to feel like ten hours of work. Three of five candidates quit partway. One of them left us an email: “There was just no way to do this while holding down a job.” They were right. A task that cannot coexist with a day job does not filter for skill. It filters for spare time.
The current task is built on three design principles. It must fit into two evenings after work, three hours in total. It must have no right answer, and ask instead how the candidate narrowed the problem down. And since we are spending three hours of the candidate’s time, we return feedback on every submission, including to those we turn down. Last month’s task was to read twenty anonymized consultation records and lay out, on a single page, what to reinforce first and why. We do not have an answer key either. Reviewing one submission and writing the feedback takes us about an hour. At our current size, that is a cost we can carry, and the day it stops being one, we will redo the math.
The Reasons We Said No
Gather up the reasons we passed on people and the pattern narrows to three. Answers that lead with solutions and never define the problem. Answers that are all success stories, with nothing given up to speak of. Answers full of “we” and empty of “I.”
We paid tuition to learn these patterns. In the fall of last year, 2025, shaken by a brilliant resume, we skipped about half of these questions and made the hire. Three months later we parted ways, hard on both sides, and the cost of those three months did not end with the salary. It took the evenings of the colleagues at Epic Prep and Playfit who covered the empty seat.
Of course, we have no certainty this math is right. Somewhere among the people we turned down for talking slowly or summarizing poorly, there was probably someone who does the work better than we do. Judging a person by three questions and three hours of homework may be arrogance in itself.
Still, for now, we are keeping this approach. Publishing the bar and being told we got it wrong is a faster way to fix it than hiding the bar and waiting for good people to come along. Putting this post out is part of that same calculation.