Haha, no worries, just wanted to add some context since people seem to like context :)

Okay. I thought I had the terms wrong, and maybe I credited the wrong person. I do really like it, whoever did it and whatever it's called. 

siggy47

Ha, you guys are both right:

- I was talking about [embeddings-based](https://platform.openai.com/docs/guides/embeddings) search, which is an incredibly flexible and powerful way to find similar content.  I still have many ideas about this -- the current [upgraded] search I believe uses something in this direction, but not what I was imagining.

- The ability to do embeddings in an embedded fashion, akin to a [PKM](https://logseq.com/) and w/ a fluid UI, would truly be a game changer.  But that's a more challenging design problem, and you could get a long way even without this.

elvismercury

The term is [text embeddings](https://cloud.google.com/vertex-ai/docs/generative-ai/model-reference/text-embeddings) fyi

k00b

Once it was implemented, the value it brings is immense in tying posts together and increasing visibility for older content.

 did to our search engine which now uses a neural network to find similar posts with "once it was implemented".

I agree, it's very valuable to find similar posts this way. Kudos to 

But I think with "embedded search", a search while writing a reply was meant.

 in your reply and a search would popup to find the post you want to link to. Just like you can currently 

 and a search for all users shows up. This way, you don't need to stop writing your comment to link to a post. We haven't implemented that yet though.

During my first year on SN I spent a lot of time dredging up old posts to read. It felt like a shame that, at least for me, they were so difficult to find.

It's still a shame that there are no obvious, only abstract rewards for that currently but we will get there, I think. :)

mostly_harmless

Capital Formation Weekly #1

> Once it was implemented, the value it brings is immense in tying posts together and increasing visibility for older content.

I think you mean the update that @k00b did to our search engine which now uses a neural network to find similar posts with "once it was implemented".

I agree, it's very valuable to find similar posts this way. Kudos to @k00b for that!

But I think with "embedded search", a search while writing a reply was meant.[^1] For example, you could simply type `/` in your reply and a search would popup to find the post you want to link to. Just like you can currently `@` and a search for all users shows up. This way, you don't need to stop writing your comment to link to a post. We haven't implemented that yet though.

> During my first year on SN I spent a lot of time dredging up old posts to read. It felt like a shame that, at least for me, they were so difficult to find.

It's still a shame that there are no obvious, only abstract rewards for that currently but we will get there, I think. :)

[^1]: https://github.com/stackernews/stacker.news/issues/542