Replying to
@robn@social.lol this actually seems like a very nice LLM use case! have it summarize *your own work*, manually vet the summary, and then publish it for the benefit of others.
i note that the vetting is presumably relatively easy and therefore less error-prone for you as the original author. and then the benefit to others is obvious, but since it's a very high-level summary, any inaccuracies that might slip through should have low impact if any.
in other words, it's a system that dampens errors by design. in contrast to putting LLMs smack in the middle of the critical path of production.
I got the first sense of this long ago when it was pointed out to me that in the interest of my being taken seriously I should not call myself a programmer but a ✨software developer, because programming meant menial labor.