Returning to this conversation I think I misread your previous comment: you are saying you did use an LLM to generate some parts of the code and text in your repository, just not all, or not all in one go, or with only a couple of prompts. Is that correct?
>> I also don't think you should be worried or disappointed about finding something interesting no matter what its provenance.
I disagree. There are two issues here: accuracy, and novelty of the contribution.
Regarding accuracy, by now everyone understands that LLMs are champion bullshitters and you can't rely on anything they generate to be factually correct. My most recent experience with that is helping a student with their MSc dissertation. The dissertation included references to four papers whose descriptions had nothing to do with the actual, published papers. I happen to know those papers well since they come from my PhD advisor and one of his students and I had studied them during my own PhD. It was clear that whatever entity had come up with the description of those papers had imagined their content based on some words in the title, e.g. one paper about learning robot strategies through higher-order abstraction and predicate invention was described as contributing a novel way to control a robot hand- completely absent from the real paper. I know the student used ChatGPT enthusiastically and it was obvious that the imaginary description of the real paper came from it. As others have argued, if we keep polluting knowledge environments (the internet, publication records) with automatically generated bullshit there will come a time when we can't tell it apart from the real content.
As to the value of such LLM-generated bullshit for learning, for the person using the LLM or those reading the generated content, it is obviously very near zero, or worse, negative. For example, it was clear to me that the student learned nothing from the four papers they cited; they probably didn't even read them. And if you were to ask them now about the content of those papers they would just repeat the LLM-generated bullshit in their dissertation. In other words: negative learning value.
Regarding novelty, if some material can be generated by an LLM, maybe with a bit of elbow grease to come up with the right prompts, then there is no real point in sharing it as an interesting piece of work- just like I wouldn't copy some code or text found on the internet in my own web page or repo, and then point to my copy as something interesting, I would also not do that for whatever comes out of an LLM. Anyone who is interested in the subject can just generate it themselves with an LLM. Maybe they can even do a better job than me with the prompts, or have access to a better LLM.
That's also one reason that I don't think it's a good use of my time to suggest improvements for the temporal logic essay on your repo, if it's basically LLM-generated (with many of your prompts as input). What's the point of doing that? The LLM won't learn anything from my suggestions. It will happily generate the same kind of essay the next time someone else calls it. With text written by a human you can at least hope that you will help them improve their future work, if you make suggestions. But an LLM?
Anyway sorry that this comment is rather more critical than my last one. I'm not accusing you of doing anything wrong, to be clear. I really advise exercising great caution when sharing work generated by an LLM, even if you feel you have contributed a substantial amount of work and are very excited about the results. At the very least you can expect skepticism of the kind expressed by other comments in this post, which I'm guessing is not the reaction you are aiming for.
Finally, if people can easily tell your code or text is generated using an LLM maybe that's an indication that it needs more work.