I believe that this kind of thinking is what differentiates the companies that are successful in the long run from the ones who can create a one shot product; and I'm not sure the OP is on the right camp.
To take an internet-famous example, Kevin Rose went on the record to say that one of the major failings at Digg was the fact that they hired anyone who could code in PHP in the beginning, but when a PHP-based solution stopped being the answer, they were left over with a bunch of PHP developers who couldn't do anything but that.
If your mexican restaurant finds out that the most successful part of the business becomes making dessert (Mexican at first, but then onto other kinds of styles), then they would be happy to have picked someone who can actually cook something else than just Mexican food.
I think the whiteboard tests are an attempt at figuring out the capacity of the interviewee to abstract the problem instead of solving it the fastest/usual way. In my experience, the ability to abstract is very closely correlated to the ability to adapt to new situations (and code better in general)
I agree that variety is nice, but the underlying thrust of this article is that programming on a whiteboard is an inferior metric to programming on a computer (at least that's what I took away). It also makes very little sense to have a Mexican restaurant that ONLY cares whether it's Chef's can cook desserts.
Whether or not you can do bit twiddling is far less relevant to a PHP coding position than being able to write PHP. This is not to say that a bit twiddling problem could be added as extra credit, but you should be far more concerned with their understanding of PHP.
So, to incorporate both of your interview styles, it would be a good idea to have the chef cook Mexican food, desserts, and some other stuff rather than just cook a single Mexican dish. This would give you a more complete view of his qualities. Then a decision can be made based on a variety of data rather than a single data point.
But flexibility can be too extreme. I've seen job postings for people who are good at programming AND can photoshop good UI. To me these are such completely unrelated fields that it would be very hard to find someone who is good at both. You would probably have better luck hiring two different people for those positions.
If your mexican restaurant finds out that the most successful part of the business becomes making dessert (Mexican at first, but then onto other kinds of styles), then they would be happy to have picked someone who can actually cook something else than just Mexican food.
Or they could give said chef the recipe and ask them to make it - something like crème brûlée would not be easily made just following a recipe but requires culinary knowledge to complete it. A proper coding test/exercise will still be able to weed people out without having to put them on the spot with a whiteboard.
To take an internet-famous example, Kevin Rose went on the record to say that one of the major failings at Digg was the fact that they hired anyone who could code in PHP in the beginning, but when a PHP-based solution stopped being the answer, they were left over with a bunch of PHP developers who couldn't do anything but that.
If your mexican restaurant finds out that the most successful part of the business becomes making dessert (Mexican at first, but then onto other kinds of styles), then they would be happy to have picked someone who can actually cook something else than just Mexican food.
I think the whiteboard tests are an attempt at figuring out the capacity of the interviewee to abstract the problem instead of solving it the fastest/usual way. In my experience, the ability to abstract is very closely correlated to the ability to adapt to new situations (and code better in general)