Proliferation of low quality software has always been the case and it's just a new thing people have latched onto now. And there are plenty of economic drivers like the whole enshitifaction phenomenon Doctorow loves to harp about that lead companies to produce garbage software. Similarly, we have stuff like the NPM ecosystem that's become a meme now. All of that was happening long before LLMs were a thing.

So, how did we figure out what projects to use? People looked at who works on the project and what their track record is, how long it's been developed, how many users it has, and so on. All of that applies to the world where we have software produced using LLMs as well. Projects that are well built and well maintained will rise to prominence, and that's where people will gravitate.

But I think there's a second aspect here as well that's even more interesting. Building software that's general purpose is a lot harder than building software for yourself that solves your own problem. Any developer knows that it's pretty easy to hack up a script or a problem to scratch your own itch, but making it public and usable for others is a ton of work on top of that. Conversely, users tend to develop their own workflows with a program and they don't actually use all the features in it.

With LLMs we're now getting to a point where people will be able to just make their own software stack to support the things they need to do with their computer. It doesn't need to be perfect or general purpose. It just needs to do what the person using the computer wants to do. I expect that's the world we might end up in. What's gonna matter are common protocols, and file formats so that people can send files to each other and communicate, but everybody will be able to have their own unique computing experience based on the way they want to use their computer.