This is of course the big problem. Facebook is admitting that developers elsewhere in the world are just as good as developers in Silicon Valley. Or if you want to put it another way: if Silicon Valley is special, it's NOT because the developers.
Now the market is crowded, scaling is a simple economic problem, the network effect is turning on its head and there isn't any revolutionary software to build on the horizon. Times have changed but the narrative hasn't caught up yet.. its going to suck when it does honestly. Software is just so commoditized and cookie cutter at this point i can't see it being a great career for much longer.
I've been trying to learn more about the dark corners of the linux kernel and graphics drivers for now. But its not 2015 and I don't have 3 years of experience in embedded/systems/C++ so no ones gonna hire me for that stuff because my resume just says "generic backend dev lol".
Software is the infrastructure of the 21st century and if you step out of the tech bubble to things like material science, biotech, etc. you'll see that their infrastructure is garbage despite there being really big returns available.
Developers who can marry modern software to those domains should remain in high demand for decades.
Anything related to public safety (including transportation) probably needs the most tech talent.
Spend a month with a tech stack of any passing complexity, even ones at $FAANGco...
We're nowhere close to solving all the hard problems.
When you already have architecture that is capable of serving the entire world, when everyone has a smartphone all that's left in that space is efficiency/cost optimization. Whatever shiny new programming language or framework comes out, the bedrock of computing is solid (and the shiny new programming language is really mashing together decades old ideas which is cool if you're new to them but boring if you're not) . Every company is a mess on the inside of course, but they do their core functions well enough. AWS did not exist 10 years ago, now it's big internet drama when it goes down for a few hours twice a YEAR.A detailed postmortem comes out (not standard before), revealing the problem was a configuration change (because everything has long since been automated, so it's driven purely by configuration) .
I know not everyone is interested in those same problems, but I thought they were interesting and what's left now in the guts of software is an adoption phase, people rediscovering lessons that others long since figured out. Perhaps people who work on higher layers of the stack think differently, I'm only interested in the deep tech below, and it's pretty solid at this point.