> I spent 10 years (even more when you account for non-profession experience) getting good at things that are becoming less and less valuable.
This is just how it is, and has always been in this industry. And it takes about 10 years to realize it.
When I started my career in software, businesses were still writing new code in COBOL. 10 years later those skills were pretty much useless, except for dwindling maintenance roles.
Then there was the client/server era. Then the web era. Then mobile. Then cloud, etc.
All the same functionality, written and re-written time and time again, using the latest popular stacks and methodologies.
I hope to be retiring in a few years and pretty much everything I have learned over nearly 40 years is no longer applicable or is at best losing relevancy to the way sofware is built today. And that's how it's always been.
Sure, but those are cherrypicked cases where a technology became obsolete. There are many counterexamples of decades-old technologies that are still actively chosen for greenfield work today, in 2026.
SQL was first released in 1973. More new SQL is being written today than ever.
C++ (1985) is the de facto standard implementation language for web browsers, JavaScript engines, networking stacks, telecommunications, video games, high speed trading, CAD/CAM, video rendering and editing, audio processing, filesystems, databases, hardware drivers, automotive, aerospace, and robotics, among others.
Is Rust making inroads? Sure, and it's a tiny fraction of C++ still. It's a long ways from being the standard.
Likewise, Python is often cited as the "AI language," but that's on the surface -- CUDA, tensor libraries, inference languages, GPU kernels, compiler stacks, and so on are usually C++.
Then there's C -- introduced in 1972. Still widely used for greenfield in kernels, device drivers, embedded systems and microcontrollers, filesystems, firmware, network stacks, cryptography, databases, compilers.
LaTeX, MATLAB, Erlang, Verilog, PostScript, Lisp (including Scheme and Clojure), shell scripting (and the UNIX paradigm itself)... the list of old tech that still sees new projects in 2026 goes on.
Honestly, even if Python had replaced C++, it was originally created in the 90s, so it's already a "decades-old" technology. Python 3.0 came out 17 years ago.
This is a good point. I started with Turbo Pascal in school and my first job was writing VB 5 Windows apps for local businesses. SQL (and C, but I haven't written it in ages) is probably the only thing I learned in the late 90s that has been a constant my entire career. Languages and frameworks seemed to change so frequently that I never thought too much about them. I was always focused on the end goal of solving some business problem.
What’s interesting is this sort of implies tech skills are inherently the most valuable part. Are they? Tech jobs probably paid the worst when they the most “technical” at the lowest level of abstraction. As they’ve become “easier”. The pay has increased dramatically.
This is just how it is, and has always been in this industry. And it takes about 10 years to realize it.
When I started my career in software, businesses were still writing new code in COBOL. 10 years later those skills were pretty much useless, except for dwindling maintenance roles.
Then there was the client/server era. Then the web era. Then mobile. Then cloud, etc.
All the same functionality, written and re-written time and time again, using the latest popular stacks and methodologies.
I hope to be retiring in a few years and pretty much everything I have learned over nearly 40 years is no longer applicable or is at best losing relevancy to the way sofware is built today. And that's how it's always been.