One thing I wish he'd stressed more: there's almost always a big gap between the research that's published (and any software used to produce it) and being usefu

One thing I wish he'd stressed more: there's almost always a big gap between the research that's published (and any software used to produce it) and being useful in a team's real-world environments. For software engineering tools, for example, researchers tend not to put a lot of time into instalability, integration into teams' toolchains and processes, usability, or clarity of error messages ... to say nothing of performance and scalability -- you don't need any of those things to publish the paper. Of course there are exceptions, and I don't mean this as a criticism of the researchers. It's just that there's a lot of work involved in "tech transfer", and it's often hard to justify the investment -- especially for a relatively-small team. Microsoft and Google and Facebook all have plenty of people doing various forms of "applied research" to bridging that gap; but they've got the resources to do it.