I find myself seeing nothing useful on Kagi, going to Google, seeing nothing useful, then asking ChatGPT and sometimes seeing something useful and othertimes being led on a wild goose chase. The general state of information retrieval seems bleak. Reddit is usually the solution for me, even on technical matters now.
Append a "?" to the end of the kagi query. It runs what appears to be a ChatGPT RAG query backed by a search engine index, and puts the results at the top of the normal search engine result page. It greatly outperforms any other LLM I've played with, and, as a bonus, each paragraph in the response contains working hyperlinks to primary sources.
If you don't want to pay Kagi or login, you can play with it here:
I didn't know it runs an LLM when you append a "?", but for any Kagi-users out there, you can use the bang: !fgpt $QUERY if you automatically want to jump to an LLM.
The !fgpt-bang seems to be the model: "Claude 3 Haiku" going by the developer notes. Which often outperforms at least ChatGPT 3.5, easily recouping some of the money I put into Kagi every month.
Search engines worked a lot better when the internet had a higher SNR in the link graph. Nowadays it's an ocean of SEO sewage and no search engine can do a good job. It's not that Google ruined search by showing ads; it has genuinely become a harder problem. There is not much that can be done except set up a federated darknet where any commercial activity is banned; otherwise, the incentives are all wrong.
this is my experience as well. sometimes I accidentally search Google and I find it extremely annoying and the results to be demonstrably worse most of the time
the example I like to show people is searching "how to fix a leaky faucet"
Kagi shows helpful answers and videos from sites like This Old House.
Google shows ads for plumbers near me. If I had wanted a plumber, I would've searched for that.
Kagi lets me raise and lower sites or even block sites, so I get results more relevant to me. If I see a site that is not useful (hello Quora) I can block them. If I see something I like, I can raise it.
These days I go chatgpt4 -> ddg -> Google. I did the Kagi trial but it wasn’t compelling.
I am generally sceptical of GPT results, but also of other results, and GPT search is easier to fine tune and drill down into. For example if it gives me an obviously wrong answer, you can call BS. And it even apologises! Much more difficult to do for search engines.
After reading the author's article and the emails, the only thing I'm convinced is that the author has an axe to grind and Vlad comes off as entirely normal and reasonable.
I find myself seeing nothing useful on Kagi, going to Google, seeing nothing useful, then asking ChatGPT and sometimes seeing something useful and othertimes being led on a wild goose chase. The general state of information retrieval seems bleak. Reddit is usually the solution for me, even on technical matters now.
Append a "?" to the end of the kagi query. It runs what appears to be a ChatGPT RAG query backed by a search engine index, and puts the results at the top of the normal search engine result page. It greatly outperforms any other LLM I've played with, and, as a bonus, each paragraph in the response contains working hyperlinks to primary sources.
If you don't want to pay Kagi or login, you can play with it here:
https://kagi.com/fastgpt
(no need to append "?" when you run queries through that form).
I didn't know it runs an LLM when you append a "?", but for any Kagi-users out there, you can use the bang: !fgpt $QUERY if you automatically want to jump to an LLM.
The !fgpt-bang seems to be the model: "Claude 3 Haiku" going by the developer notes. Which often outperforms at least ChatGPT 3.5, easily recouping some of the money I put into Kagi every month.
Thanks for the tip, I’ll give it a try
Search engines worked a lot better when the internet had a higher SNR in the link graph. Nowadays it's an ocean of SEO sewage and no search engine can do a good job. It's not that Google ruined search by showing ads; it has genuinely become a harder problem. There is not much that can be done except set up a federated darknet where any commercial activity is banned; otherwise, the incentives are all wrong.
The SEO spam domains could be blocked or de-prioritized by google. However, these SEO links generate significant income for Google via advertising.
When OKRs are tied to revenue, no executive is going to sign off on a change that reduces it
1 reply →
Strange, I rarely don't find what I need on Kagi, and when I do and fall back to Google, the results there are no more helpful.
this is my experience as well. sometimes I accidentally search Google and I find it extremely annoying and the results to be demonstrably worse most of the time
the example I like to show people is searching "how to fix a leaky faucet"
Kagi shows helpful answers and videos from sites like This Old House.
Google shows ads for plumbers near me. If I had wanted a plumber, I would've searched for that.
5 replies →
Kagi lets me raise and lower sites or even block sites, so I get results more relevant to me. If I see a site that is not useful (hello Quora) I can block them. If I see something I like, I can raise it.
I typically go google -> yandex -> kagi -> chatgpt.
These days I go chatgpt4 -> ddg -> Google. I did the Kagi trial but it wasn’t compelling.
I am generally sceptical of GPT results, but also of other results, and GPT search is easier to fine tune and drill down into. For example if it gives me an obviously wrong answer, you can call BS. And it even apologises! Much more difficult to do for search engines.
The same for me. I tried it for a month and it just didn't work well enough to make a switch :/
This describes my own experience 100% accurately.
Kagi seems to have their heads up their asses though. https://news.ycombinator.com/item?id=40011314
One dude's crappy blog doesn't change the fact that Kagi has excellent search results.
After reading the author's article and the emails, the only thing I'm convinced is that the author has an axe to grind and Vlad comes off as entirely normal and reasonable.
Agreed