Agreed. Something else that might be driving this is that existing models essentially get the job done for most users, who —- unlike HN commenters (I promise this is human generated, em dash notwithstanding ;P) —- don't quite about the state of the art.