This is pretty much the whole dilema though. By sharing it and putting it on GitHub you’re just making the AI slop contributions problem worse.

AI has also enabled me to do so many crazy projects I know are way beyond my skill level in a pre-AI world, and although I’d love to share some of them I’ve determined it’s just not worthwhile. Anyone technically inclined enough and sufficiently motivated can repeat the same process I did if they want similar tools. AI indeed makes mistakes, but it’s scary good at code and has completely revolutionized workflows in every CS position I know of in my company as well as every college CS friend’s company. It’s better at catching edge cases than humans and writes comments that actually explain what a function does in context of the rest of the program. It’s a shame it’s all built on stolen training data and forum suggestions from people who were trying to help others in good-faith, but the end result is an undeniably useful tool.

If you have any project ideas or clearly scoped changes to existing open source software you already use daily, it’s worth trying for yourself. Google Antigravity is my preferred tool for this sort of thing at a hobbyist level and has a relatively generous free tier. VS Code with copilot is more common professionally and although with the right models connected it’s generally better at outputting code that looks less like it’s vibe coded, I wouldn’t say it’s all that much more powerful of a tool for just getting a personal project working.