You're missing the adjectives. Worhtless online test, 10 hour homework, and whiteboard trivia. The point is that an interview shouldn't require a ton of a candidate's off-time outside of the interview The interview itself should be a worthwhile measure of the developers skill while pertaining to the actual job. The amount of piano tuners in Chicago is probably irrelevant to 99.9% of jobs.

That makes more sense indeed, I guess I'm just finding myself having to do a lot of adjective filtering of online comments.

Interviewing wasn't ruled out. Just all day interviewing.

That can be time consuming and error prone.

At least in Brazil, a significant amount of candidates are out of the market developers who simply are not very good.

Some of them might be good at selling themselves in an interview but will be bad performers. In many ways a bad hire will cost a lot more than no hire at all.

Similarly developers bad at selling themselves on interviews might be good performers. If they have public code that can be looked at we can decide for ourselves, but that is also hard to compare to other candidates, specially since a lot of them do not have public code.

I believe a bite sized on site (with internet)/online/homework coding task within the required skill set will actually help filter out some red flags and will help drive the following interview since it's easier to talk about something concrete.

As long as the task is close to actual work that will be done and not gimmicky, won't take an absurd amount of time, and the candidate can have access to all resources they would usually have (mainly and IDE and internet access), I think it's a valid proposition.