I think personal code will get a lot more common, too. This might actually have some benefits for security.
For example, I have been vibe coding an AI personal assistant in OCaml (like openclaw, but not shit). If openclaw has a massive vulnerability, it’s unlikely to work for mine. I don’t know if I’ll release it. While there’s a lot of code and a decent feature-set, there are dozens of these kind of projects at this point. Not to mention the effort of supporting users and all the bugs that affect them (but not me). The main benefit mine has is there’s a bunch of proofs about the code’s behavior – but what good are they if I don’t understand what they’re proving and no one ever checks? (To clarify: the proofs are valid proofs, but of what?)
I think most code in the future will be personal code. Or at least a very large chunk. Why buy a product when you can design it instead for pennies and add whatever features you want when you want? It goes for other things too that don’t need to be high quality: Why watch an average TV show when you can generate an average one instead? Why put on background music from a real artist when you can have an LLM listening to the games night you’re having with your friends and automatically generate tonally appropriate music for the moment in the story?
While there’s a lot of worry about the internet or whatever falling apart due to horrible vibe coded bugs; personal code works in the opposite direction (in some ways). There are some antifragile properties that emerge. But also if all of those vibed solutions use the same library that can be fragile again.