[carlosdp]

I don't think this is a bad thing. Pretty much all of the author's examples of "new and potentially superior technologies" are really just different flavors of developer UX for doing the same things you could do with the "old" libraries/technologies.

In a world where AI is writing the code, who cares what libraries it is using? I don't really have to touch the code that much, I just need it to work. That's the future we're headed for, at lightning speed.

[3D30497420]

The problem is if that code hasn't already been written in some form or another, then the LLM is much less effective at giving recommendations.

I've been playing around with embedded systems, specifically LoRa libraries on ESP32s. Code from LLMs is next to useless for a lot of what I'm trying to do since it is relatively niche.

[pphysch]

> In a world where AI is writing the code, who cares what libraries it is using? I don't really have to touch the code that much, I just need it to work. That's the future we're headed for, at lightning speed.

This attitude works for write-and-forget workflows where the only thing that matters is whether it returns the answer you want (AKA "hacking it").

Once you add in other concerns: security, performance, maintainability, it can fall apart.

[hypothesis]

> Once you add in other concerns: security, performance, maintainability, it can fall apart.

Does anyone care about that? E.g. CRWD is all time high, after all. There is zero need to change anything according to market.

[pphysch]

> Does anyone care about that?

Yes

[hypothesis]

That’s not much to counter my example.

[tikhonj]

> who cares what libraries it is using?

Presumably the people who have to read, debug and maintain the resulting garbage.

Then again, we have so much garbage code before LLMs, that it was clearly never that important.