Well to me it sounded like they followed the first part (ignore feature requests) and failed at the second part.
Just build the feature is the wrong lesson to learn. The right lesson is to expend more effort to understand; and yes, part of that _could_ involve shipping the feature in order to get data/feedback.
Just build the feature is the wrong lesson to learn. The right lesson is to expend more effort to understand; and yes, part of that _could_ involve shipping the feature in order to get data/feedback.