>In my essay a distribution of one I argued that bespoke software was the original, correct arrangement, that fifty years of productised general-purpose software were a compromise forced by the economics of scarce programmers, and that AI has ended the compromise.

I would also say that productised general-purpose software was an undue bonanza taking unfair advantage of copyright, which shouldn't have allowed so many restrictions on code which is necessary to make devices perform their intended tasks.

Fortunately AI has been able to get started accomplishing some of the much-needed workarounds to these annoying copyright issues, like few humans have been able to do.

Regardless of whether programmers are scarce or abundant.

On a tangent, it sounds like a famous canal, but I would figure there are a number of little-known waterways where there might be a high-performance 21st century PC resting underwater along with some poor soul's bitcoin wallet :(

The "A Distribution of One" argument seems less convincing than the "The Computer at the Bottom of a Canal" one. I'm no expert, but if you've ever read some of the computing press from the time you'll know that custom software was still rather popular well into the '80s: and unfortunately custom software written for in-house use by a non-software company was the classic venue for all the Software Crisis war stories, which probably has a lot more to do with why it went out of fashion than progammer wages on their own. Whether the main problem was the difficulty of extracting requirements from users and management or the lack of sufficiently good software developers, maybe the arrival of LLM software cannons will be what it takes to fix it, but a hae ma doots.