I've definitely noticed an uptick in output from "non-developers" specifically in game modding circles. It started when ChatGPT came out and people were copy-pasting code from it into files without knowing how to check if they actually were even in the proper file format for games to recognize, and but with tools like Claude the floor has risen to the point where most vibe-coded mods are at least loadable by games and most tooling people are writing will at least compile.

The thing that took me by surprise is that despite a huge amount of noise, there's a non-zero number of instances where the output has ended up being quite high quality. There's someone who by their own account had no development experience before AI tooling has started doing some incredible work on mod management on Linux, and from using some of the tools they've made so far, it's way higher quality than pretty much any other native Linux tooling for managing mods I've seen so far. I ended up chatting with them a bit after I got enough use out of what they made that it seemed worth subscribing to their Patreon, and they've spent a lot of time on improving their coding skills because they can tell how much it helps them even if they're utilizing the AI tooling for a lot of it. This specific developer is probably an outlier in terms of how well they're able to identify and focus on what ends up impacting the user experience (both in terms of features and willingness to spend quite a lot of effort tracking down and fixing even small bugs), but it seems like they might have never gone down this path in the first place without AI tooling.

I think the ratio of developers compared to "regular" users is still probably fairly small, so the number of people who have never had any reason to want or need a Github is still way higher than the number of people who already have accounts. I'm not trying to say that this won't cause some amount of friction in how development works or that there won't be a lot of slop out there that might make it hard to find the higher quality output, but I suspect that trying to fight against the trend is going to be a losing battle, and that "solving" it will end up requiring better ways to select which software that we want to use rather than limiting the amount of it that gets written and published. Writing software has continued to get more accessible over time; early computers weren't something you could have at home and work on stuff in your spare time on, and I'm sure that the the volume of low-quality software increased when that barrier was eliminated, but the volume of high quality software increased as well, and in the decades since then we've come up with all sorts of new ways to search for and distribute software that makes more of it available to everyone. At least right now, this is starting to look like a similar shift in how much more software is going to be produced, and although a lot of it might be bad, I'm not convinced that there won't also be some potential benefits to having a lot more people willing to try it out themselves. Trying to reap those benefits requires at least some amount of acceptance though, and I think the focus needs to be on the quality of the output independent of the process rather than on the process itself.

The bar for what we consider acceptable software shouldn't be lowered, but the litmus test that it needs to be written by a "real" developer is not necessarily going to be as accurate as those of us in the old guard would like. Yes, there will likely be problems down the line for someone who only wants to vibe and not understand what's actually going on with the code itself, but I've also worked with fantastic engineers who only have vague ideas about how a compiler works or what a CPU is actually is doing under the hood because it turns out that sometimes (not always!) that's not required knowledge to be able to write software that's actually useful. What they've all had in common is that they genuinely care about the quality of what they produce, so they focused on what the actual experience and then figured out what they needed to learn to improve that rather than starting from the bottom up.