[OpenAI] The take-home was a real engineering task, not a puzzle
OpenAI take-home surprised me: a few hours, and it felt like a slice of an actual codebase/problem, not a LeetCode set.
They cared about how you structure and reason about a real task, not just "does it pass the tests." Then the onsite goes deeper on the same muscles, so it's coherent rather than a separate hoop.
For those who did it, did you treat the take-home as make-or-break, or just a filter to clear?
Comments (3)
Sign in to comment.
Filter, in my experience, but a filter you can absolutely fail by being sloppy. Clean structure + a clear README mattered more than cramming features.
Did you get to talk through your take-home in the onsite? Mine became the basis for the deep-dive round.
Yes, exactly that. The deep-dive was basically "walk us through your take-home decisions," so write it like you'll have to defend every choice.