I should have loved electrical engineering

Which is why I also don't generally like analogies and the kind

There are plenty of "applied" electronics technician or electrician's apprenticeship programs that are more like your software education. Take an induction motor, a variable frequency drive, a few sensors, and a programmable logic controller, and hook them together according to the manufacturer's instructions, and you can be off to the races operating a pump or a conveyor on day 1. But will you understand how the insulated gate bipolar transistors and filters in that variable frequency drive turn the rectified high-voltage DC bus into three-phase AC that generates a rotating magnetic field and induces a current in the motor armature? No, you don't need to know any of that to make the pump work.

You wrote:

> I couldn’t imagine ... a toy CPU implemented in SystemVerilog being ... useful

No, it's really not, but your work on real CPUs depends on registers and combinatorial logic and ALUs and MMUs. End users can typically just download Python and treat everything behind the screen as a black box, but if you really want to call what you're doing "engineering" or a "science", then developing an understanding for what happens behind the curtains is incredibly useful. If you've implemented a toy 8-bit CPU with load, store, compare, jump, and a few basic math instructions, you can write some assembly or a toy interpreter for it and you will have an understanding of how real CPUs work that can enable you to write better code later. Add some interrupts to that CPU and build a multitasking operating system, and you'll understand parallelism better.

All of modern technology is a pyramid. At the point of that pyramid is just a single doped semiconductor with a P-N junction. We build that junction into transistors, and transistors into gates, and gates into CPUs, and on those CPUs we execute assembly, and we write low-level languages that compile into assembly, and build operating systems and syscalls with those low-level languages, and access those systems with high-level languages, and connect those computers together with networks, and write applications that operate on those networks, and at the broad base of the pyramid there are billions of people using those applications.

In 2025, no one human brain comprehends the full stack anymore, we all become our own individual bricks in a particular layer. But to do the best work at any point in the pyramid, you ought to know a bit of how it works above and below you!

I think the explosion in availability of inexpensive microcontrollers and FPGA dev boards have made it much easier for people to get into hardware design without spending a ton of money. This has also made it cheaper to buy high end test equipment, you don't need to buy a $3k Keysight oscope when a cheap Chinese USB oscope works just as well with plenty of features built-in for free. Obviously a proper academic or corporate research lab is going to be a lot different than a well-equiped hobbyist lab but the difference is not as stark as you'd imagine.

Software development for the most part is extremely easy. It's one of the few "engineering" fields where you can go to a bootcamp and learn it in 6 months. You won't see this kind of thing for quantum mechanics or electrical engineering.

Also the gap between theory and application in software is miniscule. Instantaneous even. You basically learn software via application.

A lot of software engineers take pride in their jobs and in their intelligence but they don't fully understand just how easy software is. Like you guys (to be accurate: most of you guys, not all) have an easy job and you learned a easy field. EE is challenging. You don't like it because it's harder and the intellectual capacity to handle it isn't there for everyone.

There's a reason why all hardware development moved to Asia. Software is just too attractively easy and the internet boom made it lucrative. Asians took whats available while the west took what’s most fun. And now the skill gap is more evident and we can’t go back.

For the same mental effort, you get orders of magnitude more "end product" from software than hardware, with greatly less overhead and greatly more flexibility.

Hardware is extremely punishing and "complexity friction" kicks in almost immediately. A multi-feature door alarm on a microcontroller is a one hour affair that a newbie could finagle. With a pure hardware implementation its a multi-day effort, plus another day of reworking the board to dial it in. And if you aren't copying a design, you likely need a degree as well.

There is also the fact that software pays much more than hardware, can be done remotely from just about anywhere, doesn't involve working in labs full of lead and solvents, and like the author noted, has a much higher "wow!" factor from people in general. Software makes you feel very powerful, hardware will humble you into the ground.