It might be good to define “slop”. My definition is roughly, “was created with little to no human thoughtfulness in its architecture and doesn’t have the faintest hope of being maintained in the long term”. Using that criteria, easily 95%+ of AI-coded projects I’ve encountered are slop.
For me the key thing is that a program that cannot be maintained and improved over time without introducing numerous regressions may as well have not existed, because even those that were useful at one point in time quickly cease to be so due to bugs cropping up from the project being a quivering, duct taped together mass of misery or because it stopped getting updates when the project owner found themselves painted into a corner and broke with an OS update.
Note that I am not a hardline anti-LLM type. I have my reservations, but do make use of them. I do however see a massive difference in quality between knowledgeable devs using them to augment their own abilities and non-devs or newbie devs just telling Claude to bake the whole pie.
For me the key thing is that a program that cannot be maintained and improved over time without introducing numerous regressions may as well have not existed, because even those that were useful at one point in time quickly cease to be so due to bugs cropping up from the project being a quivering, duct taped together mass of misery or because it stopped getting updates when the project owner found themselves painted into a corner and broke with an OS update.
Note that I am not a hardline anti-LLM type. I have my reservations, but do make use of them. I do however see a massive difference in quality between knowledgeable devs using them to augment their own abilities and non-devs or newbie devs just telling Claude to bake the whole pie.