I interpret it as shared state concurrency. That is, multiple threads sharing the same address space performing reads and writes on shared data: atomics, mutexes, etc. From the update I understand that he's considering things like Actors, Communicating Sequential Processes (goroutines), and data partitioning divide+conquer (MapReduce) approaches as different even if they happen to be implemented on different threads sharing the memory space.

I'm not sure on the relative prevalence but I can say that I've read relatively little about what he's calling multi-threaded programming. Mostly just a handful of lock-free data structures from the C++ guys and Rust's take on how they can do it while avoiding data races due to the type system. The alternatives, on the other hand, have been on HN pretty much continuously since HN started up.