It seems quite obvious as surely the amount of C++, C, etc. is less, in the public domain, than the millions of lines of JS, C#, Java, Python & others that are available on github, gitlab & others.

So for those users, you might still be finding LLMs lacking. It's also understandable that you might get more errors when generating C++ than say C#, since it's harder to use.

It reminds me that I recently changed companies in part because at my previous job, I didn't have as many opportunities to work with LLMs (doing the things they are good at) as my current one.