Identical experience. They had me design and implement a practical thing. It was an interesting problem, not something I would normally be able to do in 30 minutes. They wanted it in 30 minutes. After about 20 I stopped coding on the whiteboard and started explaining where I was going with it. It was not complicated, just using a hash to store things, but the removal involved moving used blocks a free list of pre-used slots so the overhead of adding a new costly element was way cheaper. This would beat a binary tree for the scale of elements they where talking about but still easily kept in ram. I am sure they didn't have a clue what I was talking about. I think they wanted an RB tree based algorithm and a more primitive solution. Right after the interview I implemented it and sent it to them. I didn't even get a reply. About two months later I got a survey. This was after three interviews. The happy ending is I have used that bit of code several times since and it rocks. :)