2 ms·
This is a very cynical take (I'm cranky from my second vaccine dose), but: Imagine how much work, how much energy and effort, went into building an ML-based im
by dmitrybrant 5y ago
This is a very cynical take (I'm cranky from my second vaccine dose), but:
Imagine how much work, how much energy and effort, went into building an ML-based image cropping feature, just because an up-and-coming Designer decreed that a certain specific crop ratio was the most aesthetically pleasing to the human eye...
...so that years later, after countless hours of additional user research, they would just remove the feature because it doesn't work, and award themselves a medal for doing it.
- paxys 5y agoIt’s often the opposite at “hot” tech companies going through a massive hiring/growth phase (which I imagine Twitter was back when this feature was implemented). You hire one too many highly paid experts and then realize there’s no real practical work for them which couldn’t be done by a new college grad, so the end result is needless over-engineering.
- snowwrestler 5y agoThey address this directly in the article. The previous horizontal crop was not an arbitrary design decision, it was put in place to improve the browsing of tweets by fitting more of them into the viewport. Their intent was to improve the user experience. The ML-powered auto-crop was the engineering solution for how to deliver the best version of that. They don’t say it in this article but I suspect the need to fit more tweets into the viewport was obviated by smart phone UI. The screen is tall and it is super easy to scroll quickly with the flick gesture. No need for short images anymore.