Sure, BFS is 'easy' but it's not really something that stays fresh in your memory after years of not having to use it.
I completely agree with his point that it if you have to study for an interview, the process is broken.
Yes?
The point of an CS education is just as much to study well known algorithms and re-apply them to new contexts as it is to invent new ones. In fact, for most professionals, I'd argue the weighting is well slanted towards reapplication.
Our dev team previously used FizzBuzz, and found that every applicant was passing it. So they wanted a new take home programming problem, and internally their most important criteria was 'demonstrates algorithmic thinking.' They mostly write and maintain simple Django webapps, so I thought this was a bit silly, and suggested if they wanted to measure that, give candidates a problem easily solvable with topological sort. Unfortunately they bit hard and only found out afterwards that pretty much no applicants solved the problem correctly. We still hired four people from that round. So in some sense, turnover within a year will mean that none of our dev team can pass their own interview.
OTOH, I totally agree with the conclusion of the article: long interviews of technically complicated stuff is just BS. The last time I had to do this, I started refusing to answer questions after about two hours.