I think something like Copilot would be great if it could keep all or even just related notes in it's context.
> making connections between notes > related notes in it's context.
Mem -> similar mems
Obsidian -> https://github.com/brianpetro/obsidian-smart-connections | https://github.com/deepfates/silicon
> asking questions/searching
Mem -> search is NLP/AI by default
Markdown -> https://github.com/debanjum/khoj
Obsidian -> https://twitter.com/Sarah_A_Bentley/status/16110695760993362...
> new (summary) notes based on (many) old notes
There are a lot of summarizers on the web. They work great on whole articles. The problem is, how do you summarize hundreds (and more) of independent/related smaller notes ?
Type out a list of things you need to do in code:
1. We need to accept batch applications not just one
1. To accommodate this, we need a new parameter in the Gizmo accepting function
2. The REST API needs an optional new parameter to accept an array of applications. The code in this function also needs a new error if the array is populated as well as the old parameter for the single application.
With the zero shot learning of chat GPT, when I get home I'm able to drop them in and it'll spit out a polished version of them. It has been saving me hours of time.
Just released an integration with OpenAI's GPT3.5 (not to be confused with the ChatGPT product) which allows you to do all sorts of really cool things with your notes like summarize, argue against an idea, or for it, etc. Basically anything you can do with GPT is accessible directly in your notes.
Also it's E2E encrypted so your notes aren't the product, which I really like.
Right now I am using it as a smart news reader with a single class "do i want to see more articles about this?" It works, but I don't think this kind of content-based filtering is going to be competitive with collaborative filtering for most people and why this is not a product on the market despite having been technically possible since the early 00's.
I am thinking about extending it to take an ontology and learns models to automatically put tags on documents, images, things like that and more generally set up some workflow where it learns how to perform certain steps on its own.