Show HN: ThreadQuilt: AI-Free Thread Aggregator

> I built ThreadQuilt because I was tired of wading through endless noise to find meaningful discussions. With ThreadQuilt, you get a cleaner, more focused experience that highlights the best parts of the web's conversations.

I applaud your effort.

I’ll ask the obvious question around the strongly opinionated approach of not leveraging “AI”. Would you like to expand on what that means to you? I would imagine that some form of intelligence could also help with noise reduction but maybe you considered it and decided otherwise.

To be honest, I would rather use AI, because it would be more fun and cooler to work with. However, I have a track record of making things no one wants, so I really wanted to focus on what I thought would be more popular. Given I keep hearing that AI has "ruined" Google and other platforms, I figured it would be a good decision to not only cut it out from this but to emphasize it as a "marketing" strategy.

I am absolutely not opposed to it, and if I can convince myself that integrating AI would be moreso "what people want" then I would integrate it (and have more fun).

“AI” (as absurdly broad of a term as it is) has legitimate use case. It isn’t JUST hype. However, because of the buzz, it’s being shoehorned into so many places it really just isn’t the proper fit for, and it’s hard to figure it where.

However, there ARE plenty of areas it IS the right fit for. Lots of “fuzzy” systems that would struggle to be rule-based and generalizable benefit hugely from LLMs and other fuzzy / intentionally broadly scoped tooling.

Source — I work at a “chat with your data” startup, and our product just categorically wouldn’t be worthwhile if the above weren’t true :)

This is really cool, but I would need the dates with the results.. if I'm looking up something technical, dates really matter as libraries change etc.

2 questions... why AI free? Is it only pulling from reddit, hacker news and stack exchange? I find a a lot of tech discussion has moved to discord, which makes it very hard to find... just even looking at your search samples, kubernetes and react both have official discord servers, and i'm sure there's talk about them in countless other servers.

I can implement Discord threads! If anyone has any suggestions of what other threads to aggregate, let me know. The main issue I have to overcome is latency getting too high when retrieving form different places at once.

Additionally, I chose to avoid AI as I have picked up from a lot of people that Google’s AI features have more or less ruined it. Do you think there is a way that I could use AI here that wouldn’t ruin things?

You may have to start building your own reverse index if you're polling API's at query time

"The main issue I have to overcome is latency getting too high when retrieving form different places at once." The solution for that is maintaining and updating an index. (Although that drives up costs.)

Google has done a terrible job, but perplexity has done a pretty good job. I think there's quite a bit to be gained by summing/extracting desired information from threads. I'm thinking opinions on choosing tech stacks (eg, "should i use sql or nosql", "fastAPI vs supabase"...), or even fixing bugs (on that note, if you're going after tech, github might be another to look at).

I defitley agree there's a ton of noise in discussions, I think LLM's are the current best way to cut through that noise.

I had no intention of making a search engine, I really just envisioned a place to combine threads from different platforms and figured out along the way that it looked exactly like search.

This seems to be an overcrowded market so I would be afraid to just "make a search engine". It seems difficult to me at the moment how exactly I can make ThreadQuilt a unique value proposition if I take that route.

If AI is the differentiator, then I would absolutely want to integrate it.

This is cool. I clicked "deep learning models" and the results were ok, but some of them were out of date. It would be great if all of the threads were recent (within the last week or so).

Just an FYI,

https://www.threadquilt.com/search?topic=openbsd

Throws a 500 and displays a stack trace of the server.

I am retrieving threads from APIs per query, and stackexchange is rate limiting me. This is frustrating but a sign that search is popular! I will work on increasing rates with them, and long-term I have to educate myself on indexing.

Looks like you're getting hugged to death right now. Looking forward to checking this out in a few days once things settle around

Question: Should I combine threads and label them by origin instead of separating them like I have currently?

If anyone has any feature suggestions, please let me know. I am trying to create as much value as I can here.

It would be great to have a sorting feature by date, popularity, # comments, etc.

I’ve made some major updates everyone. Please let me know what you think.

At first glance this looks excellent! I'm going to keep playing with it.