The tech industry has "eaten its young." By this, I mean that it has systematically destroyed the source of people who grow up with the passion to learn tech a

The tech industry has "eaten its young." By this, I mean that it has systematically destroyed the source of people who grow up with the passion to learn tech at the deep levels that make for the people the tech industry likes to hire for the core, foundational stuff that makes things run. About a decade ago, I understand Google's SRE hiring pipeline was struggling, because Google had singlehandedly destroyed their best source for hiring - university student sysadmins. Google Apps for Education was great, and allowed shutting down a ton of annoyingly expensive server rooms at universities - but it also meant that those universities no longer had nearly the number of various student sysadmins who were entirely unfazed by "The mainframe is serial linked to the G3 Mac for ethernet access, with software written by someone long since graduated, and..." sort of systems. It was a wonderful filter for people with a diverse set of experience across a range of "You did what to the what now? In assembly?" sort of systems, and they could usually translate that experience to the sort of "haven't seen this before..." stuff that Google ran at the time. The same seems to be true in the really foundational parts of computing. I know very, very few people under the age of about 40 in the firmware/osdev/hypervisor/etc spaces, and that age has more or less been tracking along with me. It's not a base age for people getting into the space, it's the age of the youngest of the last generation to know how to do that stuff. There are exceptions, but they're few, far between, and seem to not go into those spaces terribly often either. The modern tech industry is just focused on distracting people as much as possible, as long as possible, to look at as many ads as possible, and that's entirely at odds with the requirements to get good at the deep weeds, which is long, sustained focus, in challenging and arcane spaces. But it's worse, because the complexity of modern tech is increasing at such a rapid rate that you have to run to stand still. All the microarchitectural vulnerabilities and such have been "patched" with... complexity. When they stem from complexity. And the whole stack at this point can be summed up as "Complexity created to hide the problems caused by the previous round of complexity." It's nasty, all the way down, and then you discover that the people who are designing the foundations have no idea what they're doing either! Intel clearly doesn't understand their chips, given how many times the stated guarantees of SGX (that malicious ring 0 code can neither observe nor modify the state of production SGX enclave computation) have been violated by a range of mechanisms. L1TF/Foreshadow cracked production enclaves wide open with demonstrations of how to single step them and pull full state from every step, and Plundervolt involved undervolting the chip to the edge of faulting (so multiply and AES operations would fault but simpler stuff would retire fine), and then launching an SGX enclave that ran faulting operations. Whoops. I don't know how much longer this sort of thing can continue, because the last wave of people who do understand the deep weeds that make it work are starting to retire, with nobody to replace them. And they just want to go do "something that has nothing to do with computers" anymore. A recent comic on Reddit (https://www.reddit.com/r/ProgrammerHumor/comments/vnmmuo/where_do_you_see_yourself_in_5_years/ https://www.reddit.com/r/ProgrammerHumor/comments/vnmmuo/whe...) demonstrates just how strong this sentiment has become in recent years. The hypervisor hackers and firmware devs are going to retire and rapidly forget everything they've known, and have a grand old time doing "anything else." I know an awful lot of these people, myself included. It's going to be a very real problem in the decades to come.