- The task was to implement a basic tic tac toe AI (with open book !!!)
the state space is so small that anything would work fast enough in a modern computer.
- Some candidates somehow manage to build a program that loses?? that is, it's buggy and they failed to test it.
- The interviewer uses subjective criteria that somehow results in broken code being accepted unanimously.
- Some people overengineered their solutions and this is called impressive.
- Someone failed to implement the exercise, wrote an answer that can be googled in a couple of seconds. Made an irrelevant comment about "min-max" vs "negmax" ( performance is not a concern (small state space) and negmax's main advantage is being DRY).
Made a comment about "test bloat" despite empirical evidence from the submissions showing that there was under testing.
and the conclusion that I am supposed to have is that interviewing without code is better? a worth-mentioning portion of the candidates couldn't ship a working implementation of a standard algorithm and the conclusion is less code??
To be honest tic tac toe is a bad task to choose. it's trivial, easily googlable, and has little room for skill expression. It's also doesn't test the candidates' ability to choose the correct algorithm for the task. and it fails to probe daily aspects of dev work ( concurrency, networking, databases ..) that traditional onsite interviews are less able to test.