How cool is Cuil?
Posted by Andrew Myhre on July 29, 2008
So there’s buzz around this new search engine, but I want to know if it’s good, and whether it’s something I’d actually use. Off-hand I can think of three types of search I perform reguarly: technology, social networking and Wikipedia. Technology searching demonstrates up-to-date indexing, and the latter two demonstrate a level of index depth. So I performed a couple of tests, and here are my results with handy links if you want to compare for yourself.
Note: Cuil will probably improve a lot, and these are three very specific searches so we can only draw limited conclusions from them. So take all of this with a pinch of salt. If you know of a good comparison search let me know and I’ll add it here.
Search Results
Technology: Search term = ‘asp.net mvc preview 4′
Cuil: http://www.cuil.com/search?q=asp.net%20mvc%20preview%204&sl=long
The first page gives me eleven links to content, none of which is specifically to do with the latest build of the ASP.Net MVC Framework, let alone a link to a download page. Some of the pages are tutorials relevant to older versions of the framework but may still be applicable to the new version.
The striking thing to notice in Google’s results page is that, of the ten results, nine of them actually have the words ‘preview 4′ in the title. From this I can immediately gather that their results are more relevant to me. Also, three of the pages contained links to the MVC project page or the installer on CodePlex, and one of the pages was a direct link to the CodePlex project page.
Social networking: Search term = ‘twitter andrew myhre’
Cuil: http://www.cuil.com/search?q=twitter%20andrew%20myhre&sl=long
Interestingly, Cuil’s first page provides a list of people I follow/who follow me, but none of the results is a link to me or my Twitter page. Most likely that’s because noone would deliberately link to my feed, but those links do exist on the pages in Cuil’s search result.
Google gives a somewhat better result. The first link is to a tweet within the feed of someone who follows me, and the second is a link to their profile. The third and fourth links are to my FriendFeed and the fifth is to a much-less-up-to-date mirror of this blog. The rest are barely or not at all related to me. Somewhat better than Cuil but I still expected better somehow.
Most interesting is when I search using my actual Twitter username ‘andrew_myhre’:
Google finds me, first hit.
http://www.cuil.com/search?q=twitter%20andrew_myhre&sl=long
Cuil actually returns zero results. Disappointing.
Wikipedia: Search term = ‘wiki bicycle pump’
Google: http://www.google.co.uk/search?hl=en&q=wiki+bicycle+pump&btnG=Google+Search&meta=
As expected, the first result is the exact record I’m looking for – the Wikipedia entry on bicycle pumps.
Cuil: http://www.cuil.com/search?q=wiki%20bicycle%20pump&sl=long
Cuil earnestly presents a number of options but comes up short. No links to Wikipedia to be found.
Summary
Generally, Cuil doesn’t give me information that’s as relevant as what Google does. The problem seems to lie in the relationships between content nodes – why can’t Cuil knit my blog and social networking sites together to provide a seemingly complete ‘whole’ in the way that Google does?
On top of this, the value-adds that Google provides like postcode parsing, shopping information and natural language parsing of terms like time zones, weather and money (interesting that they don’t parse Zimbabwe dollars btw) I just get a lot more out of Google’s results.
