The Miracle Interview — cover art: a small rabbit at a Victorian whiteboard, paw on chin, mid-recursion, with an older fox in a tweed jacket gesturing from a chair — the rabbit's drawing splitting open a hidden symmetry under chalk diagrams.

2005 · Drunken Blog Rants · Rant

“He'd gone from screwing up nearly all my questions to acing nearly all of them. He started off needing a ton of help and requiring prompting for almost everything, and wound up deducing the correct meanings from vague questions with lightning speed.”
— From The Miracle Interview, March 2005
Read the essay

© 2005 Steve Yegge. Originally published at Drunken Blog Rants.

AI Notes

A March 2005 interview write-up, posted publicly with identifying details removed. The candidate — an SDE-2 with an M.S. and two years of experience — opens by failing Steve's regex/grep question (proposing to write a five-day C program instead), fumbles string permutations because he can't see the recursion, and makes the small but telling mistake of scrunching his work into the rightmost two inches of an empty whiteboard. Steve does something he rarely does: stops the interview, writes the bullet points of the failure on the board, and tells the candidate plainly that this isn't going well. The reversal follows. The candidate listens, nods, and proceeds to smoke a Parking Lot OO design, the entire threads/processes/IPC battery, and — given an arithmetic tree, the eval-function shape, and a single nudge toward polymorphism — derives the Plus/Minus node-class solution out loud in under a minute. The candidate knew the material the whole time; what failed early on was how he metabolized the interview.

Steve treats the loop as a controlled experiment in feedback latency. Most interviewers withhold honest signal until the debrief; saying the quiet part out loud at minute twenty unblocked a candidate who had been throttled by his read of the situation rather than any gap in his grasp of the problem. Postscript, in Steve's own footnote: everyone else on the loop voted not-inclined. Miracle cancelled, I guess.

Related listings

  • 2005

    Five Essential Phone Screen Questions

    The companion methodology piece. The Five Essential Questions are the very pre-flight checks that, applied to this candidate, would have routed him onto a phone screen instead of a plane ticket.

  • 2005

    Why Phone Screens Matter

    The sibling case study, two months earlier. Same loop shape — failed every question — but with no miraculous reversal. The two essays bracket the same point: the pre-flight check is what's missing.

  • 2006

    The Truth About Interviewing

    A year later on Blogspot, Steve writes the broader theory of interviewing of which this essay is one suggestive data point — including the part where direct, honest feedback in the room may produce better signal than the polite version.