> The ACM SIGPLAN talk is a very good starting place to get an understanding what he is talking about.

What's the big difference from just using composition like we've done for ages?

I mean, his examples in that talk was very reminiscent of stuff I've been doing for many years, and I don't consider myself to be special in any way. Writing decent interfaces, write useful implementations, then compose them together.

Not trying to be dismissive, I just feel like I'm clearly missing something as it's a bit underwhelming for a 2019 talk to be all about "composition is great, use it".