The state of the art of software development does NOT move fast, it moves glacially slowly. The fundamental technologies of 20 years ago are very similar to what is in use today. The tools are similar, the languages are similar, the frameworks and libraries are similar. Mostly, they have simply now more polished and have a few more features.
A handful of things that were leading edge 20 years ago have gained popularity. These include things like object frameworks, event-oriented programming and the like. Some things that some call new are actually things that existed in the 80's that have just come back into fashion. Take asynchronous programming using callbacks (popular with javascript/node). We used these techniques in the late 80's quite extensively. There are a few things on the leading edge today that did not exist decades ago (or at least I wasn't aware of them). But these are far and few between. The basics though, those have not changed.
What changes rapidly is tech fashion (for want of a better phrase).
When I hire, I simply do not care if you know framework X or Y. Or even if you know language P, Q or R. What I care about is if you 1) truly understand the fundamentals and if you can 2) apply that knowledge to create high quality software.
Don't master the 'surface' technology, master the underlying concepts. For example, rather than trying to master this week's cool machine learning library, master the underlying mathematics and you'll not only be far ahead of most people who only mastered the library, but you'll happily be able to transfer your skills to the next machine learning library with minimal effort for the rest of your career.
I've heard this many times before, but have not seen any evidence yet. Mind listing the technologies that got really, totally obsolete?
So often I see the next great thing hype turn out to be a reinvention of an existing concept, or a slight rearrangement of concepts that are positively ancient. Sometimes this includes an incremental improvement over the past.
A lot of the velocity of "tech moves fast" is just backfilling tooling, features, concepts, and edge cases into the new language or framework.
Here I am now working entirely with the AWS infrastructure. First we shoved things in Puppet. Then we started shoving things in Git and using Ansible. Maybe development wise things don't seem that different but for an old ops guy like me having to go devops it is changing very fast.
I never said that old technologies are obsolete but, jobs for supporting them fade and we must learn to survive in the new world.
When hiring I usually looked for guys that were smart and could quickly learn new things or jacks of all trades. Now if trying to write new kernel drivers maybe I'd get a Kernel/C expert who values himself at 120k/year (probably contract so I could use him for a month or two). But, lets face it most of the time companies want to hire the 3 smart enough guys at 80k/year than 2 geniuses at one specific language or technology. At least that has been my experience, then again, I usually work at normal smaller/midsize and non tech large companies. I'm sure if you work at Google your experience would be very different.
The point is I don't care about being the expert anymore as it requires a lot of my unpaid personal time. Then after investing my time for free I find the tech trending down and have to go back to the drawing board again.