I've worked in a bunch of different parts of the stack over the course of my career. One thing I found is that often the "hard" work of previous decades is lar

I've worked in a bunch of different parts of the stack over the course of my career. One thing I found is that often the "hard" work of previous decades is largely encapsulated, sometimes it's pretty difficult to make further incremental gains. I'm currently working on a large performance critical C/C++ backend code base, which unfortunately has the problem that most of the bugs are related to memory management. I'd be hard pressed to decide whether all of the defensive coding we've added over the years carries more overhead than a bump-allocated tracing GC or not. There are hard problems on the frontier, Electron makes it possible to create one app which runs on every platform and every modality. How we manage the complexity of hundreds of different UIs in one app is still largely unsolved (but getting better). Pretty soon we'll have non-JS lang's compiling to WASM as a standard method of developing UI. Managing dependencies which were built to target a POSIX system in WASM is still unsolved, but once we solve it - it'll be possible to do all sorts of crazy things like shipping libtorch to the browser :)