Despite years of predictions that AI would kill coding interviews, the DSA (data structures & algorithms) format has stayed remarkably consistent at big tech companies.

  • Coding has changed dramatically, but the interview format is "the one area that have surprisingly stayed pretty consistent" — companies are only experimenting mildly with AI-assisted interviews 3:19
  • DSA interviews were never meant to test job-relevant skills; they evaluate whether "somebody can think" 4:23
  • Second reason: companies have no idea how to evaluate candidates — and probably never did. Amazon ran internal studies showing that no matter how much interview data you collect, predicting on-the-job performance is very hard 4:46
  • Even a technically good hire may fail for reasons interviews can't measure: motivation, team fit, environment 5:09

Why change is so slow at big companies

  • "If it ain't broke, don't fix it": any change risks making the process worse 5:18
  • Changing the process is bureaucratic and requires retraining thousands of interviewers — most of whom hate interviewing. Meta's AI coding interview experiments show how hard the training is 5:32
  • Big companies need a standardized process; new evaluation formats add variables (AI prompts to check, different training) that make consistency "practically impossible" to measure 5:49
  • Prediction: new formats will appear, but the transition will be much slower than most people expect 6:15