Commenting two minutes after the Show HN was created suggests the author asked friends to come to HN to comment. I'd like to point to the guidelines "Please don't ask friends to upvote or comment. That's not ok on HN." https://news.ycombinator.com/showhn.html
Hmm, I can't see the submission on https://news.ycombinator.com/show yet. Could be a sign most upvotes were not counted or the flagged-comment (brand new account who immediately praise how great something is) to upvote ratio is too high now. Best option is to submit again in a couple of hours, same title and text. Try first comment "author/maintainer/creator/team here, happy to answer questions" or a variation, that often works.
You take Notion as an example and I couldn't agree more.
We are their competitor and followed your exact line of thought trying and not releasing GPT in editor to avoid noise, and building ASK, an engine powered by semantic search, GPT (and a lot of other fun models for refinements), to answer all questions you could have on your knowledge base.
It's private, fast, works with real time edited documents, and it's already in production for thousands.
We tackle this hard with slite.com, we just introduced slite.com/ask to find answer no matter how badly organised your documents are, and we offer document status to track outdated and up to date docs across your space (and we're still enhancing it, very much first version).
These among many others to help it not become a behemoth as you say.
We built the technology in house and don’t rely on chatgpt, but GPT alone, and services offer it in a similar way algolia or elastic offer search for instance, it keeps data private.
We just released a new feature for Slite, Ask, in private beta.
Slite is a team workspace that hosts all kinds of knowledge: wiki, docs, projects, discussions, and decisions. When we started playing with ChatGPT, we really wanted to offer a similar experience, by basing answers on your team documents.
It felt like a parallel change of paradigm: ChatGPT to get answers from the web, Ask to get answers from your internal knowledge.
The technology under the hood is of course different: It uses GPT3 and advanced semantic search, with special ponderation based on your team activity, document linkage and so on. And of course it respects privacy - its answers are only based on what you can access.
The interface is less conversational - for now. But it's already a massive upgrade from your classic search experience, and it goes beyond that. You can really ask it anything, even stuff that's not specifically documented. The engine will look at various sources and compile them to answer.
First testers love it, and we believe it can radically change the way we work with documentation in the future. Check it out!
We just released a new feature for Slite, Ask, in private beta.
Slite is a team workspace that hosts all kinds of knowledge: wiki, docs, projects, discussions, and decisions.
When we started playing with ChatGPT, we really wanted to offer a similar experience, by basing answers on your team documents.
It felt like a parallel change of paradigm: ChatGPT to get answers from the web, Ask to get answers from your internal knowledge.
The technology under the hood is of course different: It uses GPT3 and advanced semantic search, with special ponderation based on your team activity, document linkage and so on. And of course it respects privacy - its answers are only based on what you can access.
The interface is less conversational - for now. But it's already a massive upgrade from your classic search experience, and it goes beyond that. You can really ask it anything, even stuff that's not specifically documented. The engine will look at various sources and compile them to answer.
First testers love it, and we believe it can radically change the way we work with documentation in the future. Check it out!
Thanks for sharing! What was inferior about Slite (and when did you use it)?
I’m working on it, would love to hear your takes, especially as in term of performance and speed we worked heavily this year and believe to have reached something superior to alternatives.
I wish I had access to the issues I was going back and forth with your team on, but I no longer have access to that email account.
One big thing we wanted was relationships. Perhaps we’d keep data in a table and some entries would link to another document. At the time, this wasn’t possible but I was told it was on the horizon/being considered.
Otherwise we had issues with performance slow downs when we allowed documents to get very large. This was generally bearable, but we had some documents we didn’t want to break apart (for reasons), and opening them up became a little frustrating. They didn’t scroll well and editing them was laggy, for example.
I trust your team has been working hard on it because it was very evident that the team cared a lot and was super attentive. I constantly worried I was being annoying, but I got quick, steady, and very gracious responses, always looking to provide useful information. Super positive experience in that regard!
I can't edit this now, but it just occurred to me that the relational data issue was specific to Slite, not Notion. Racking my brain, I'm fairly sure the issue we had most with Notion was performance.
The performance definitely is something we improved a lot, and indeed relational data - I guess you mean roll ups in database- is not something we allow at the moment, thanks for sharing again!
If you like performance and have a team usage, Slite should be an interesting choice to explore.
Disclaimer I'm the founder, hundreds of team come from notion every month for this reason. Our search is much more powerful and lightning fast, navigation and doc loading is super fast as well, we put extra care there.
