RE: https://neuromatch.social/@jonny/116326861737478342

This thread is really interesting, and raises an issue which might be some weird corollary to Moore's law. Anyone who was programming before, I dunno, let's pick 2001, remembers the shift from constrained-resource coding, where you *had* to pay attention to resource use, make your algorithms as efficient as possible, etc. We didn't all manage to code cleanly and concisely, but that was the goal. 1/3

Then we started noticing that as resources became cheaper, and available RAM and HD increased, corporate programs began to bloat. Fast forward 25 years, and we have this Claude Code mess, where you can just spin up inefficient agents, and call wasteful JSON validation whenever/however you want, because resource use no longer matters *at all*. 2/3
And this weirdly goes along with the increased cost of e.g. data centres, RAM, etc, not because software "really" needs these resources, but because Gen AI is a method of burning capital. Moore's Law ends up becoming the computing equivalent of burning the peaches in The Grapes of Wrath. 3/3

@redlibrarian I remember a prof of mine in 2001 saying she had to write code with paging (ie making sure it could fit in RAM) and we all lost our minds in class.

The C Union datatype is another blast from the past as well. I couldn't imagine life needing to write code like that.

@elibtronic @redlibrarian now I'm having bad flashbacks to the perl internals
@elibtronic when I was in comp sci in 95-96 you *had* to know the sizes of your data types and how to optimize at the variable level. I never had to do paging myself but I remember learning about it.
@redlibrarian @elibtronic you guys oughta do programming for microcontrollers - that'll take you back.
@adr @redlibrarian Kids these days don't know how good they have it.