Given that people have access to LLMs themselves, publishing their output in lieu of good documentation (no matter how sparse) seems like it’s mostly downside.