alt Hacker News> No one is going to write new UI libraries if SOTA models know React best, no one is going to bother with new languages if SOTA models know Python, Go, JavaScript, and so on the best.
Eh, you just don’t get how people work. You might be right overall, or maybe not.
People will write new UI frameworks and languages if the problem seems interesting to them. There are loads of people who just like working on interesting problems, and that’s very unique to each person. They may use AI, they may not, but as long as people get deeply sucked into interesting problems, we’ll still see new ideas and projects.
Or, look at react. It was invented because Facebook was running into issues writing big web UIs. Many companies still have issues writing these, and if they have issues, it’ll be a way they can differentiate and compete. What happens when someone turns an AI on the optimization problem and invents a bespoke in-house tool that’s actually fantastic? That’s another way you could get new tools.
Replies
"People will write new UI frameworks and languages if the problem seems interesting to them."
I would think even with AI assistance industrial quality solutions (which is what we are talking here I think ) still require corporate size sponsor to thrive.
Yes, anyone will create whatever suits their fancy. But there is huge gap between hobby project and industrial adoption.
Yesterday, I wrote down a list of UI libraries for myself.
There are many, for every web design framework (for React/etc or PHP).
So yes, people write new libraries. The question is how they’re going to write them in the next 5 years.
Couple of aspects you are not considering.
1. If majority of developers use LLMs for coding, and then if someone developed a better React, then no one can really use it because the models does not know how to write idiomatic code in it. It might be easier to make the LLM implement workaround for the issue with the current React, than making it use an unfamiliar library or paradigm.
3. Without some critical adoption rate, the new library won't have enough training material, and this becomes a chicken and egg problem. This can only be broken by AI companies by somehow training the models on the new React by code explicitly written, but again, without critical adoption, there is no incentive for them to do it.
Imagine we are living in the time of jquery and all the LLMs are proficient in it. I don't think in such a world, no one would want Angular/React, because LLMs will happily handle all the complexity that using jquery results in.
We would have stuck with JQuery forever..