Why do you think there is a lot of training data? Could it be because it's stable and virtually unchanged for decades? Hmmm.
Because bash is everywhere. Stability is a separate concern. And we know this because LLMs routinely generate deprecated code for libraries that change a lot.
alt Hacker News
Because bash is everywhere. Stability is a separate concern. And we know this because LLMs routinely generate deprecated code for libraries that change a lot.