It may be an interesting side effect that people stop so gratuitously inventing random new software languages and frameowrks because the LLMs don't know about it. I know I'm already leaning towards tech that the LLM can work well with, simply because being able to ask the LLM to solve 90% of the problem outweighs any marginal advantage using a slightly better language or framework offers. Fro example, I dislike Python as a language pretty intensely, but I can't deny that the LLMs are significantly better in Python than many other languages.
It kinda makes me sad. I hope we don't enter an era of stagnation because LLMs only understand our current tech. There are so many brilliant ideas that mainstream languages haven't adopted (and may never be able to adopt). There are language features outside of python and JavaScript that will change the way you think about systems and problem solving. Effects systems, structured concurrency, advanced type systems, distributed computing... There are brilliant people discovering safer and smarter ways to solve these problems but it seems nobody will care anymore.
Yes, sticking to the most popular technologies increases quality of the output, enabling even smaller startups to build applications like https://youtu.be/oafdA2WXvEc?feature=shared
Alternatively, esoteric languages and frameworks will become even more lucrative ,simply because only the person who invented them and their hardcore following will understand half of it.
Obviously, not a given, but not unreasonable given what we have seen historically.
why would it be lucrative? The person paying would consider whether they'd get locked in to the framework/language, and be held hostage by the creator(s). This is friction to adoption. So LLMs will make popular, corporate backed languages/frameworks even more popular and drown out the small ones.
Scarcity of some knowledge. Not all knowledge exists on SO. You are right about the popular stuff, but the niche stuff will be like everything else niche, harder to get and thus more expensive. COBOL is typically used as an example of this, but COBOL was at least documented. I am saying this, because, while I completely buy that there will be executives who will attempt to do this, it won't be so easy to rewrite it all like Jassy from Amazon claims ( or more accurately, it will be easy, but with exciting new ways for ATMs, airlines and so on to mess with one's day).
<< The person paying would consider whether they'd get locked in
I want to believe a rational actor would do that. It makes sense. On the other hand, companies and people sign contracts that lock them in all the time to all sorts of things and for myriad of reasons including, but not limited to being wined and dined.
Again, I think you are right about the trend ( as it will exacerbate already existing issues ), but wrong about the end result.
I created a new framework and fed my documentation + certain important code snippets into it. It worked out fantastic. Now adays though, the LLM never follows links and will hallucinate the whole thing, in a completely wrong language.
You can probably fine-tune a general purpose programming model on the code and documentation of your language project (the documentation being in large part written by an LLM too, of course).
