Thoughts on LLMs

Something that LLMs are really good at is explaining things, at least at an intermediate or basic level. For example, if someone told me that they want to learn Node.js I would tell them to forget Google and go straight to ChatGPT; you won’t be bothered by blog spam or ads (at least for now) and you can have it explain exactly what you want to know.

With classic tutorials, if you didn’t understand the original author’s explanation you’re out of luck—with an LLM, you can just ask it for a different explanation; it’s like having the author right next to you.

(This is not to say that I don’t think there’s no value left in educational content, but I feel like long-form educational content will have the same relationship to LLMs as books to do the Web)

However, that raises the question—if people stop writing tutorials, where should the LLMs source new knowledge from? Publishing manuals for new tech might be enough; however, I suspect that people will just skip the middleman and start training the LLMs directly. I suspect that training LLMs will play the same role as producing documentation does today, and become an integral step of releasing a product .