Continuation of `Having "blessed" 3rd party libraries may make D more popular` DIP thread

[GrimMaple]

On Tuesday, 8 July 2025 at 21:30:24 UTC, WraithGlade wrote:
> Stealing from many people in aggregate and then using 
> transformations to hide the nature of that (a.k.a. "training 
> the AI") is even worse than stealing directly the old fashioned 
> way. [..]

I'm sorry, but in this regard AIs "learn" in no different way 
than us humans do; realistically speaking, you are able to 
program mostly because you "stole from many people in aggregate 
and then used transformations to hide it". Most of code online is 
licensed under permissive licenses anyway, and many of that code 
is MIT which is literally "do what you want".

> This is a matter of basic human rights, not "politics". [..]

I'm yet again sorry, but this is a matter of ridiculously toxic 
intellectual property rights laws, that has infested and 
sabotaged human development for years now. When you develop 
something physical, like a windscreen wiper, you're most commonly 
allowed to hold the patent for no longer than 20 years. But when 
it comes to IP for some reason it can be more than a hundred 
years. Maybe if Microsoft wasn't allowed to close-source their SW 
forever we'd be living in a better place, who knows.

LLMs have been a quality step up in terms of technology, and they 
are a great help already. However, it seems that there is a 
physical barrier to how good they can become, and we'd need 
another quality step up in computation to make them smarter/more 
affordable.

I understand how people are discouraged by LLMs because they feel 
like they're gonna "take their jobs", but on practices LLMs are 
nowhere near good enough to actually replace a programmer, or 
anyone else - it's just a tool that can spit out some generic 
code that pretty much any CS intern can do. Except it would take 
a student a few hours to do so, and an LLM can do it in a minute. 
So the programmer can concentrate on something actually important 
instead of spending time to re-invent stuff that has been 
invented a hundred times already.

I remember adr made a point that code reuse is a myth only to get 
laughed at. However, if you look at that in a more broad and 
philosophical way, you could easily see that most of the code you 
write is indeed something that has already been written. And your 
code re-use rate is incredibly small, aside from some basic 
ground up 3rd party (like .NET MVC).

For decades programmers have been taking pride on their algorithm 
and structure knowledge, and here comes LLM that can do all that, 
but much more effective. And all that prized knowledge is 
suddenly useless. But in reality that algorithm knowledge has 
been fairly useless for more than 10 years now, it was just 
fairly hard to accept for most programmers. But now they're 
forced to focus on something useful, and not the code itself.

In summary, that's my entire point - code doesn't matter. 
Software does. So if you focus on creating software rather than 
writing code, you might eventually find joy in both programming 
itself and maybe even using LLMs for it :)