I think the big reason I dislike almost all discussion about AI is that the content is pushed as a way to help you do something with less effort, as if the only thing holding you back from writing that amazing novel you have in you, is the time it would take.

It couldn't be that, you don't write at all and thus are a bad writer who has bad ideas because you've never practised the craft of writing. It couldn't be that people are in love with the idea of "having written" as in the idea that the work is done and people love what they wrote and they didn't have to do the hard part in the middle.

Oh wait, that's exactly what AI is all about. Convincing use that some tool will make good ideas easier. It will simplify book structure, and make that blog post advertising our work appeal to everyone in a way that we never could on our own.

The truth is, that good work takes time, time few people are willing to spend. The single recurring piece of advice I give to people when I coach them on their note process is to put aside time to work their notes. No one I've coached has ever had time set aside to simply take notes and think. They don't schedule time to summarise articles and make connections. They barely have time scheduled to read the source materials they need to read.

If I wanted to run a marathon, I would get on some sort of training plan so that I could accomplish the big goal of a marathon. It would be foolish, and obviously so, to assume that some external tool would make my body ready to run the marathon with no work on my end up front. If I told someone that I had some special custom robot legs that would let me run the marathon without any effort, no one would think that I had run the marathon. My robot legs ran the marathon and I did nothing.

There is little different in letting AI do the work. You did essentially nothing, the AI did everything and likely did it poorly.

That doesn't mean I don't see useful things possible with current LLM models. The Write of Passage Podcast recently did an episode1 that highlighted the use of NotebookLM. The ability for NotebookLM to be given a paragraph of text and then find other relevant content in your notes is interesting. I have around 8000 notes now and I'm sure I've forgotten some of them. If Obsidian could take a look at some of my text and suggest other relevant ideas for me to investigate, that would be a decent use of LLM/AI.

Of course I'd still have to put the ideas together and then fit it into the overall structure of my content.