> and to date there seems to be no appetite for legal enforcement of these obligations.

Three paragraphs earlier, the virtues of openly sharing code were being sung.

> while if I keep my code closed it is far more difficult to find those same mistakes

I doubt that is significant for frontier models. They can disassemble and won't need readable code to understand what your program does.

The feeling of not wanting your code to be gobbled up by AI is something I can sympathize with, but now think it's misguided: it's just accelerated - your code doesn't depend on popularity as a metric to be adopted - and really not that different from sharing with a MIT license, as long as it doesn't up gated behind closed models and walled gardens.