[abraxas]

> You can’t survive a steady erosion of mastery.

That sounds like an onset of a certain type of dark age. Eventually the shiny bits will too fall off when the underlying foundation crumbles. It would be massively ironic if the age of the "electronic brains" brought about the demise of technological advancement.

[MagicMoonlight]

Just look at current software.

Windows is maintained by morons, and gets shitter every year.

Linux is still written by a couple of people.

Once people like that die, nobody will know how to write operating systems. I certainly couldn’t remake Linux. There’s no way anyone born after 2000 could, their brains are mush.

All software is just shit piled on top of shit. Backends in JavaScript, interfaces which use an entire web browser behind the scenes…

Eventually you’ll have lead engineers at Apple who don’t know what computers really are anymore, but just keep trying to slop more JavaScript in layer 15 of their OS.

[com2kid]

I was once one of the mush brained morons hired to work at Microsoft.

I think I did ok. Would I compare myself to the greats? No. But plenty of my coworkers stacked up to the best who'd ever worked at the company.

Do I think MS has given up on pure technical excellence? Yes, they used to be one of the hardest tech companies to get a job at, with one of the most grueling interview gauntlets and an incredibly high rejection rate. But they were also one of only a handful of companies even trying to solve hard problems, and every engineer there was working on those hard problems.

Now they need a lot of engineers to just keep services working. Debugging assembly isn't a daily part of the average engineer's day to day anymore.

There are still pockets solving hard problems, but it isn't a near universal anymore.

Google is arguably the same way, they used to only hire PhDs from top tier schools. I didn't even bother applying when I graduated because they weren't going to give a bachelor degree graduate from a state school a call back.

All that said, Google has plenty of OS engineers. Microsoft has people who know how to debug ACPI tables. The problem of those companies don't necessarily value those employees as much anymore.

> I certainly couldn’t remake Linux

Go to the os dev wiki. Try to make your own small OS. You might surprise yourself.

I sure as hell surprised myself when Microsoft put me on a team in charge of designing a new embedded runtime.

Stare at the wall looking scared for a few days then get over it and make something amazing.

[RealityVoid]

> I certainly couldn’t remake Linux. There’s no way anyone born after 2000 could, their brains are mush.

This is certainly false. There are plenty of young people that are incredibly talented. I worked with some of them. And you can probably name some from the open source projects you follow.

[piperswe]

Young people's brains have always been mush, according to the older generation. Your brain is mush according to those older than you. The term for this is juvenoia, and it's as old as humanity.

[samiv]

Nah this isn't right. We also have access to a ton of information even regarding arcane things such as writing x86 boot sequence in real mode or writing boot loaders. More now than ever before.

In fact today on GitHub alone you can find hobbyist OSs that are far far more advanced what Linuses little weekend turd ever was originally.

Their success is not gated by technical aspects.

[holoduke]

You should go outside of the "web" world. Automotive, medical or heavy industries. You will see that their are plenty of low level developers/engineers our there. Yes even ones born after 2000.

[oytis]

> Linux is still written by a couple of people.

How is that? It's easily the software project with the largest number of contributors ever (I don't know if it's true, but it could be true).

[anon291]

I've worked with many people born post-2000 who could write an operating system kernel. Hell, I have one brewing righ now. It's not rocket science. The machine language parsed by the chip is described in exquisite detail in any processor manual.

[HoldOnAMinute]

Windows is being deliberately enshittified by rent-seekers.

Rent-seeking and Promo-seeking is the only motivation for the people with the power.

None of that class wants to make a better product, or make life better or easier for the people.

[noduerme]

I'm personally not comfortable with the widespread argument (not yours) that moving from a shit-on-shit scripting language to using an LLM to write code is just a matter of degree or just one more layer of abstraction. I think writing code with an LLM severs the logic chain you may have to follow - and hopefully can follow - to get back to the basic principles and root causes you need to understanding what your code is doing.

But the lazy (and wrong) belief by people not committed to exacting standards in their engineering, that AI is just another layer of abstraction or another scripting language, actually obscures a much more unpleasant fact: Performance, as far as the managerial class was concerned, was never about getting the best performance. It was always about whatever was just enough.

We as coders used to prioritize performance because hardware was so limited and we wanted to squeeze the most out of every cycle, every 1Kb of RAM. For some of us, that habit will never die, because we look at a new piece of hardware and realize how much more we can make it do.

But pre-AI slop of backends with huge supply chains and Electron as a frontend arose because memory and compute had become so cheap that acceptable performance required less and less optimization.

That doesn't mean that some of us didn't maintain a niche in making things optimized, but for the past twenty years or so there's been a whole generation of engineers whose priority has been speed of development. And from the perspective of a company that treats engineers as disposable cogs and prioritizes frameworks and assumes Moore's Law, why not?

AI just takes that to the next level. Take the entire chain of existing React slop and create a Markov chain to regirgitate parts of it on cue. And let's be honest: 95% of companies don't need to forge anything particularly new, they just need to cobble existing parts together.

I thought it about 15 years ago talking to CRUD coders who hated their jobs: You're in the wrong business if you're not getting joy out of creating and solving new problems. So in a way, AI just gives everyone who only wanted shitty software the shitty software they deserve? I don't know. I haven't finished thinking about it.

[mooreds]

If you like dark SF, I'd suggest Pump Six by Paolo Bacigalupi. It's one of a collection of short stories, but addresses this very idea.