Compilers and high-level languages assuredly reduced the number of engineers needed, right? No: we just started making much more complicated things and the basic level of "this isn't even a serviceable solution" has gone up so high that if you sat around and wrote a trivial program designed to be used over a teletype--which is got much easier as you can do it in C instead of assembly--you'd lose in the market to your competitors who are willing to put in the effort to build a modern fancy reactive single-page whatever's-hot-now web app.

The reality is that companies are going to continue allocating a similar percentage of their budget to tech that they were before, and what we DO might change a ton; but, frankly, I've been doing this now for decades, and the stuff I used to do 25 years ago (omg: a full quarter of a century of professional software development ;P) already often just isn't a thing anyone does anymore. Every time my job gets easier, the expectations for the products I architect also increase, because we're all in a rat race attempting to compete for the same customers to defend our margins.

Hell: I remember when the first step to deploying a new service--even if you were a company that only had a handful of people--was to attempt to predict how much capacity you needed and then buy a ton of parts so you'd be able to sit around and build machines for your server room to run your new product. I was involved in a project back in like, 1998/1999, for a company with all of maybe five people, and someone's job was just sitting around building the server room... that specific task is now a few lines of YAML configuration and a couple of accounts (AWS and Twilio <- back in the day, we had to have custom T1 lines brought in not for data but for bulk parallel phone service to hook up to the Dialogic boards we needed to simulate phone lines ;P).

The concept of cloud-computing entirely changed that part of the business--having a small handful of companies manage all the servers at scale and run numerous jobs on shared hardware, in some sense reducing the cost to the point where you can both outsource this and pay less--to the point where doing that almost sounds insane; and yet, I'm sure there are more people getting paid more to deploy and manage ever more virtualized servers in the cloud than there ever were people (including me!) literally screwing around with screwdrivers in the ubiquitous server rooms. You could argue that my job got "replaced", with all of the skills I had to deal with IRQ balancing and automated power backup systems being for naught... but, somehow, I'm apparently still relevant.

I'll also admit that I honestly kind of enjoyed doing that work?... and yet, I also was quite happy to never have to do it ever again. I'll never forget the day I took the first real-world server I was in charge of and just like, uploaded it to 1&1 where I had an early software-based hypervisor-like stack I was experimenting with (UML, aka User Mode Linux, which is no longer something people bother with due to the advent of Xen/KVM/etc.); it just worked and I was like "omg I'm never building another server again, am I? that was it: it's over". (It then turned out that I did build another complex hardware solution in 2016 to start doing work with GPUs, but there was only a small window of time there before cloud services caught up.)

So like, sure: at some point the AI is going to entirely replace what we do--not the specific frustrating coding tasks we do today, but the entire concept of our jobs--but, the issue at that point isn't going to be me figuring out what I'm going to do next to make ends meat... it is going to be that I'm not a serviceable warrior in the war against the machines as the world devolves into some kind of Terminator-inspired hellscape; because, if you can manage to do my job in its totality, then you frankly can also do the job of everyone who has ever hired me and you can do the job of every investor who has ever given us money to do those things... we are all obsolete and, if AI actually gets as good as so many people here fear it will, at best we all become pets :(.

> companies are going to continue allocating a similar percentage of their budget to tech

I think your argument requires a paradigm shift like teletype -> web to soak up the new productivity. I don't think we have that today. For how many purposes have we reached peak consumer software capability? i.e. global reach to publish text, images, video and make purchases is the entire use case of most media and commerce businesses. If you gave every company double the amount of dev resources tomorrow, do they even have anything to build? Even for those with deep feature roadmaps, how much of that stuff is just nice-to-have things that won't move the revenue needle anyway?

Maybe we will get that new platform from AR or direct brain interfaces... but without it, the situation looks more like how the human role of "computer" just vanished... sure we then got programmers, but they weren't necessarily the same people.

Would being the pet of an AI-run civilisation be so bad though? Iain M Banks's Culture novels explore basically this concept, and his vision of utopia is quite appealing.

Is Jernau Morat Gurgeh the Player of Games, or a mere pawn moved by Minds beyond his comprehension?

The Culture appeals to me, but there are literary critics who find even The Culture, let alone the grander environment beyond it, to be dystopian. And I definitely don't want to wake up dead with my mind in one of Joiler Veppers' simulations.

He's definitely a pawn. But I'd sign up for being kept around as a pawn/pet/whatever in an instant if it meant living in the Culture.

I honestly don't like how we treat our pets (and especially so here in the United States). I keep imagining that the situation will end up feeling--at best--like this episode of (The New) Outer Limits. :(

https://theouterlimits.fandom.com/wiki/The_Human_Operators

>Would being the pet of an AI-run civilisation be so bad though?

Only for people with dignity. So, in general, no!