Hacker News new | past | comments | ask | show | jobs | submit login
Show HN: Future of Niche Search Engines (kampadre.ch)
2 points by rstnpce 9 months ago | hide | past | favorite | 2 comments
“Google is slowly dying.” is a phrase I hear surprisingly often these days and I think there is some truth to it. Yet, it is barely the whole truth.

That being said, I wonder why this is, if it's true at all and also what the future of search engines—or gathering information in general—will look like…

By now we all know that Perplexity AI, ChatGPT and all the other LLMs are able to answer your questions just like Google does (oftentimes even better, without the unnecessary clutter).

But why is it that I still sometimes default to Google and its archaic keyword search instead of the conversational, human way of looking up stuff?

IMO there are three reasons:

1) It’s faster. 2) It does the job. 3) It doesn’t require you to interact with it like a human(!)

Let’s zoom in on the 3rd point first, because I think this is the most insightful one.

To do so, I want you to think about your last interaction with an LLM…

I’m sure something like this happened:

You made an effort to say “please” and “thank you”.

Probably you have also thought about how to prompt your chatbot in a way that “it” didn’t feel stressed out in anticipation of a superior result (“imagine yourself in a calm and focused state”). Or maybe you have even looked up the best ways to prompt your LLM to get the best result (very meta…).

I can assure you, you are not alone and this happens to me all the time. I know ChatGPT has a temper more often than not and it doesn’t feel a 100% like a machine anymore (everything is a little uncertain)…

To be honest with you, I don’t think this is the UX that we all want. It’s just something that we put up with in return for intel that we wouldn’t get otherwise.

With simple info, however, we would never want to have these thoughts nor the delay in the answer. We just need an accurate (2) and quick result (1) (e.g., “capital of switzerland” => “bern”).

Something that good ol’ Google Search does exceptionally well.

But with the “Google is dying” premise defied, the big question remains: What will the future of search engines look like?

Also, how are keyword search and LLMs handling requests that are in-between super simple and highly complex? Will it be a smart combination of both or something entirely different?

In a shameless thrill of the shill, I’m presenting a prototype of a “niche search engine” that might partially answer this question.

Kampadre allows you to search in a keyword-like fashion with quick results but AI in the background to understand what you truly want.

The way it works is that you first describe your (marketing) goals (keywords or full sentence) (e.g., “i want to have more sales of product x online”). Afterwards, the request gets translated into choosing between a predefined set of filters which then outputs your desired result in no time.

Like this, 'smartness', accuracy, and quickness are in balance for the optimal UX. If you wanna take a look you can try it under the link provided.

Of course, this is just one way future search engines for everything in between highly complex and easy information could handle requests and serve results improved by AI.

But I’m sure you have other ideas or suggestions. Thanks for reading until here!




> By now we all know that Perplexity AI, ChatGPT and all the other LLMs are able to answer your questions just like Google does (oftentimes even better, without the unnecessary clutter).

nope they haven't been able to, the responses are slower, worse quality, incorrect, outdated and gives me a long-winded answer instead of giving me a short and direct answer

> You made an effort to say “please” and “thank you”.

nope i havent

> Probably you have also thought about how to prompt your chatbot in a way that “it” didn’t feel stressed out in anticipation of a superior result (“imagine yourself in a calm and focused state”). Or maybe you have even looked up the best ways to prompt your LLM to get the best result (very meta…).

nope i havent


> nope they haven't been able to, the responses are slower, worse quality, incorrect, outdated and gives me a long-winded answer instead of giving me a short and direct answer

Exactly what I'm saying: For simple stuff, google is still superior. But for complex stuff (i.e., code snippets in PHP). ChatGPT is way better.

Perplexity quotes it's responses, ChatGPT now does so too in many cases – if they're wrong, so is a google search since the source quality is bad.

There's outdated data by design, a point that i missed, i agree. I hope the automatic fallback to bing search is getting better (faster).

> nope i havent

I assume you haven't tried to use these LLMs to their full potential then?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: