my take is that either it IS copyrightable and thus every single fucking piece of LLM-written code is a copyright violation of biblical proportions (because it's trained on everything), or it isn't copyrightable and thus effectively unusable for open source things