Hacker News Reader

Comment by dmotz | Hacker News Reader

I built a tool for myself for the purpose of grokking ideas from books called Emdash [1]. Over the years I've collected reams of highlights from books and articles but until recently, rarely reviewed or absorbed them. The core of this app uses on-device ML to show related passages with similar ideas from other books you've read, and I find that going broad and exploring concepts from different angles really helps in comprehension.

I'm testing out a summarization/rephrase feature backed by LLMs that you can try in the demo. In HN fashion I'm trying to build this openly and gather feedback to see what works. I'd like to push this further in the active direction the article mentions with something like a Socratic dialogue mode where you're nudged to re-explain and examine ideas.

If anyone uses this thing/has feedback, let me know. Source is available too [2].

[1] https://emdash.ai

[2] https://github.com/dmotz/emdash

View on HN

This is cool. I imagine instructors creating instances for their classes so the whole class can engage with each other's notes.

I wonder if one could couple it with OCR so that you could point a phone at a page and drop into an emdash experience on the text that you've got a physical copy of. Or, you know, point it at your kindle so that your notes aren't locked into their ecosystem.

I'm building a backend that would support that kind of thing in a peer to peer kind of way (indexes content by piecewise hash so that you can recognize content you or your peers have annotations for and reattach those annotations despite differences in pagination, etc). If I ever get it into a demo-worthy state, I may reach out to see if we can make them work together.

Yes, I'd like to focus on more seamless ways to import your text and OCR is something I've considered.

Your content addressable system sounds very interesting, let me know when you have a demo.

This is really neat!! Love how you have a sample library for one to experience how using emdash is like.

This is such a neat tool. The presentation is very pleasant. Is the intention to have the snippets/notes be shareable in the future? I actually made a similar tool [1] (though your's is much more complete), which I use to quickly find passages and relevant text when I'm blogging. And, I was thinking it might be really useful to have highly rated notes on a snippet available so that you can get someone else's insight on a particular selection. I'll give this a more in depth look later when I want to write another blog post.

[1]: https://ishmael.app

Thanks for sharing, I like the name and I'll try it out. Yes, I'd like to add opt-in sharing features in the future -- seeing others' notes can be very insightful as you said.

I tried 'related' with a couple of passages. For the first passage, the first result was a good semantic match, but the rest were a little too far off. For the second passage, the results were amazing.

Perhaps for the first you just didn't have any more snippets that were closer?

Are the related snippets taken from a selection of snippets you created, or from the full text of other books?

A nice workflow might be to select a passage I'm reading in a book, and then see related passages from other books. But that requires I have DRM-free ebooks, and that these have already been chunked and indexed.

Yes, the demo mode is a random subset of things I've highlighted and it's heavily weighted around certain topics and sparse on others, so that's why some passages don't have the strongest semantic matches.

You're right that it would be nice to see things in situ as you're reading, but it would seem that most e-reading experiences are locked down. I appreciate the feedback!

im really enjoying reading your samples. It's a very lightweight way to explore books i don't know about and haven't read.

i like it a lot!

I’ve built a similar thing at https://findsight.ai

The design is beautiful! I'm shocked this isn't more popular on github.

I've been looking for something like this to review books. Two suggestions: 1. allow to load a book from a url, so notes could be added to arbitrary books; 2. allow to select text and add notes to that selection.

I would really love for something like this to be integrated into Logseq

Reply on news.ycombinator.com