But perhaps I'm tainted because I find Netflix to be annoying, on the whole. Crappy playback options, iffy streaming of HD (even when torrents have no trouble), inconsistent audio and subtitles, and an atrocious recommendation system that seems aimed at wasting my time than recommending good content. More and more I find myself using Popcorn Time, even if the content is on Netflix.
These are goalpoast moving requirements, but I also wasn't the person making it out to be trivial.
Also, all these things are "good ideas", but can you imagine going to your boss and saying. "I have three-9s of the cases covered, and for the vast majority we simply refuse to show slides and almost nobody notices ... But ... I need another six months of tweaking the facial detection to get to four-9s before we ship it. To make sure nobody seems a funny face when they pause the video."
Have you considered what the expected revenue differential between the two systems is (with and without funny-face detection) and have you compared that to the cost of the time spent in a meeting discussing it, let alone the work to implement it, and then to the opportunity cost of not doing other stuff?
These comments about potential difficulties are bike-shedding in action. Nobody has anything to add so they add nothing. I'm sure we could all imagine cases where there were complex restrictions, and that'd be totally cool if these was an article about rule-engines. But it's not.
What other product offers that? How is it central to their value proposition; what customer left because ...
Really? it takes in a movie and spits your thumbnails, sans spoilers, sans family un-friendly content, into your data store, already indexed? Across multiple data centers?
Seems like you're making it out to be much more trivial than it is.
EDIT: To be clear, the feature itself has trivial components. But rolling out a new feature requires far, far more than just the frontend code and an ffmpeg script.