Naturally we are not discussing a small programmer making their fast clean code at home or a small studio producing in house software, not sure anyone actually suggested that.

I think we know who we are talking about here, the likes of nVidia and Microslop who both currently use AI to "Vibe" code their output, who both use massive monolithic libraries that have dozens of dependencies that make even the smallest of programs require gigabytes of both ram and drive space. The big companies that release a dozen updates that are broken because their own programming team have no idea what the AI code is actually doing or what's in the libraries and thus need to rewrite the code to make it work correctly.

These are who we are discussing here, teams that should know better but continue to follow lazy practises. I'm not blaming AI here, Claude Code can be incredibly useful but AI output is based on the input from the user, great input and a knowledgeable user will produce code far superior to a lazy programmer just wanting to get it done quickly and thinking the AI can do that for them.

In the end its not about reinventing the wheel its about knowing what a wheel is to begin with and how it works and if a random wheel off the shelf is the best option for your use case.