Ex-Google engineers debut 'Cuil' way to search | Hacker News Reader

Ex-Google engineers debut 'Cuil' way to search

news.yahoo.com

81 pointssnewe107 comments

And now TechCrunch:

http://www.techcrunch.com/2008/07/27/google-beats-cuil-hands-down-in-size-and-relevance-but-that-isnt-the-whole-story/

Open articleView on HN

At least Cuil has some respect for privacy:

"We do not keep logs of our users’ search activity."

"We analyze the Web, not our users. Read our Privacy Policy for details. It’s short."

http://www.cuil.com/info/privacy/

But Jo(e) User doesn't really care about that, he or she cares about search results. If (and admittedly this is unproven) Google can deliver better search results by logging and analyzing search activity to personalize results, then Cuil is already at a disadvantage.

How can you call it unproven? When I google for Python I get stuff related to programming because of my search history, when a zoologist looks for Pythons he'll see stuff related to snakes.

Because I don't know what algorithms Cuil are using, I can't say at this stage if they are able to deliver better search results than Google without personalization. Perhaps they're better at contextualization for example, and don't need to know what you did last time.

I'm a Python programmer too and the first link I get from Google for 'python' is a Wikipedia article on the Delphic Oracle... Admittedly I have often searched for content relating to Oracle databases!

Context is always helpful and obviously should be used in addition to personal history and even override it, but if the query has little context, the only context you can get is from the users historical data.

Personalization doesn't guarantee better results but it obviously improves results.

I'm not saying that's not true, just that Cuil must believe that they don't need it, and it's too soon to see if they're right.

I do most of my searches on www.msdewey.com anyway :-)

Historical data may as well be misleading because the profile of a user is not "static", simply because there's a human being behind it (and sometimes even more than one, but that's another story). His personal and professional environment, his wishes and necessities, the context of his life may change from one day to the other. The assumption that the past queries of a user will improve the results of his current queries is at least questionable (especially for former programmers that start selling reptiles...)

Weird, I've searched python once or twice (never really used it) in the past, and don't get a Wikipedia article first, I get http://www.python.org, then the whole page is filled with python (programming) results.

I've never really thought about this, but I'm kinda glad Google logs my results, it helps me out tons by giving me results I want. Plus thier motto is "Don't be Evil"...

I search for python pretty regularly on Google. My recent, logged-in, searches for 'python gecko' resulted in a bunch of reptile sellers.

I search for programming-related things, and all but two results (both videos) for "python gecko" are about Mozilla bindings for Python.

Basically, I am not upset that Google profiles me. I would get upset when my profile is admissible as evidence against me in court, but that's not up to Google. Let's not make searching the web difficult because our legal system is fucked up. Let's fix the legal system.

I think we're probably getting the same page - I get:

* YouTube - Leopard gecko and royal python eat mice!!

* FS: 2 captive-born Mali Uromastyx - via auction (gargoyle python gecko boa) Reptiles. ...

before I get:

* Re: [pyxpcom] Python web crawler using Gecko :: ASPN Mail Archive ...

I'd have thought, given my history, the reptile versions would come later.

BTW: your comment comes up 6th on my google search for 'python gecko'.

When logged out, the search results are slightly differently ordered.

Exactly. I've seen some Google ads that were a bit too targeted lately -- related to stuff that I searched for last week -- and it is getting spooky. I'd like to have a decent alternative. Maybe I should just get used to Yahoo, but this looks hipper, no?

That being said, they haven't indexed my site yet. Amateurs!

That's very worrying. Adverts people might be interested in shown to them? Cause for concern...

You're probably intentionally missing the point -- but what I was referring to is the fact that they are quite unashamed about retaining data about me.

"The browser with this cookie, did these searches previously"

I don't think this is worrying. If they start using personal data they hold, sure, but this is just creating a personalized experience. Lots of sites do it. If you don't like them doing it, disable cookies for them...

I don't think that's worrying. If Amazon or Facebook or any other site did it wouldn't even be noticed. I think that's just reaching for a way to fault Google.

BTW, an interesting application of personalisation is not so much search ads (They are already pretty much the most targeted ads because search term is enough to target ads with) but content ads (eg adsense). Currently, ads are targeted using relevance to page content/site content or demographics of the users. But search history reveals a lot about interests & could be the key to 'fixing' context advertising (from the advertisers perspective).

what I find shady is that its against the law to spoof your competitors domain name, but google will allow you to display ads for when people search for your competitor.

thats a little underhanded if you ask me

I believe you can ask them to remove ads on particular terms, if you hold a trademark on them. I recall reading about a lawsuit that covered that issue several years ago, and I think Google lost. But I could be imagining legal precedents again.

Not exactly the same thing. Spoofing you comp's trade name is pretty much the definition of trademark violation.

It (Google's policy) flops around a bit. It differs between countries and seems to get tweaked or turned over every so often. I believe in the US bidding restrictions on trademarked terms were recently removed. But you are still restricted in using trademarked terms in ads.

Personally, I agree that restrictions (bidding on trademarked terms) should be removed. I would compare it to handing out flyers on the street outside you competitions' shop. Tasteless maybe, but shouldn't be illegal. Companies don't own the search results, trademark or not any more then they own the street.

This means they miss big opportunities to improve their search results.

"We're working on a new subway system. We're going to base its layout on building patterns rather than where our riders actually go. It would be a violation of privacy, and just plain wrong to check which buses are crowded."

Definitely not ready for prime time.

I did several searches that said there were multiple pages of results but clicking the link to the second page brought up "no results".

I'm also not sure how fresh their index is. I went to a page from the results that didn't have the content the search said it did. Normally there would be several reasons for this but I happen to have enough knowledge of the target site to know they just have a very old version of it indexed.

Trying to like this but failing. It didn't get even decent results for any of the queries I fed it.

I wonder if competing head-on with Google is a wise strategy. It's really hard to beat them without years of effort, and until then no one wants to use your site. There must be some other way.

Same for me. I tried two test queries and had none of the results that I wanted to see (which I do see on Google).

My name gets 2,290 hits on Google and 0 on Cuil, and many of those 2,290 are on some of the biggest domains on the Web. I'm having trouble believing Cuil's claim of indexing more pages.

It seems to me that a way to potentially creep up on google would be to make a specialist search engine. Some topic area that google is not so useful for searching on. Make a search system that 'understands' the domain in question, allowing it to index and search it really effectively.

As a random example, say, a movies search engine. If it could be consistently more useful than google for finding out information about movies, directors, actors, the movie industry etc., then it could become the 'go to' tool for movie search. Potentially, this success could then be leveraged to move into other areas of search.

Well, you're pretty much describing "vertical" search engines. There are a lot of those these days, and I think they have good chances of being successful, too.

It's like what we're doing over at bug.gd-- building up an error-based search engine. Our upcoming version (we hope to announce in the next month or so) will have a variety of techniques optimized for searching for errors on the web. (E.g. filtering of user-specific "noise" in error messages, better integration into software, etc.) With the added bonus of capturing feedback from users that can be used for future users and for software writers.

We've been seeing some great uptake lately, so I'm bullish on the vertical search space.

We've been debating on a two-column search results page for our redesign, so we're watching cuil's success closely to see how people like that kind of presentation.

What you should make is a specialist search engine for hackers. They're the early adopters in search engines, so if you get them you have the quickest path to getting everyone else.

Imagine how alarmed Google would be if there were a different search engine the top 10k hackers all used, even if they had everyone else. And rightly so.

http://news.ycombinator.com/apply ...

That's optimistic, but to be successful with "hackers" the search engine would most likely have to steer its user experience in directions that wouldn't appeal to an everyday user. Typically, us geeks want both maximum control and elegant usability, always preferring control to usability. With most users, the inverse is true.

Sure, there are obstacles, but my instincts tell me a clever person could find their way around them. Or at least, considering the size of the prize, the odds of finding a solution makes the problem worth taking a shot at.

Heh. I'm not amongst the top 10k, but I've been cursing yesterday, trying to use search to find out in what version / gem of ruby {|[]| } is a valid block (with just being random variable names). And a week before that the same problem when trying to find out what &@/ ment in a mathematica script.

Search engines tend to ignore symbols in normal search, and while you can search for symbols in google's code search, that doesn't include any documentation or pages where they explain a particular syntax.

Indeed. Good luck finding out information for the common shell <<END feature if you don't know that's called a "here document".

you mean like octopart?

imdb for your movie example.

I think that's how Google made some distance in the early days. I was using it and telling my friends about it when everyone was still using Yahoo!

This is exactly what Microsoft is doing. Microsoft is trying to tackle search in different verticals: healthvault for health, Farecast (acquired) for travel, Cashback for commerce, etc.

The challenge is that users don't always think "hmm, I'm doing a movie query, let me go to imdb". They go to Google and type everything into one box, and expect to get a link to imdb or wherever.

Well isn't that the point of Yahoo BOSS? To enable people to make their own vertical web search engines?

I searched for one of my own websites and was extremely pissed to find that it found a bunch of more popular sites that link to it, but not the site itself. Eff that. I'm 100 percent in favor of google competitors, but that's just invalid...

Three major problems that I see right off the bat:

1.) Quality of Search-- In order to be successful, the quality of their search must be at least as good as Google (obviously it much more helpful if it's better). As many have noted, this clearly is not the case right now, and slightly complex queries such as "Linux search and remove multiple files" return zero results.

2.) Revenue Stream-- As noted above, it is going to be difficult for Cuil to break into the search market because they have to achieve such a high quality of search. But even if they are successful breaking into the market, how are they going to generate revenue? An Adsense like program? That's just another thing that they would have to do better then Google, and I just don't see both things happening.

3.) Overcoming Inertia-- Even if Cuil addresses 1 and 2, the moment of inertia of an average Google user is very high. Google has the best web mail available on the market, and along with all the other services Google provides, people will be reluctant to switch their searching to Cuil because everything else that they do will still be through Google.

I don't mean to be negative, and more power to these guys for taking on an area of the web that most people are afraid to even look at. But I guess I would say that if I were to plan a startup, I would probably pick a more advantageous slice of the market (probably one that doesn't exist at the time of planning). At the same time, you have to play your strengths as a programmer, and it seems like thats what the Cuil team is doing.

One big part of quality of search is figuring out which sites people click on when they search for something and making that result more relevant. I can't imagine Cuil not doing that, and over time their results will probably get a lot better. Considering they started off with no data like this, not bad so far.

Anything that can compete with Google is a good thing. I have already added them as my default search in Firefox and hoping they get better. Results are not yet perfect yet, but at least they are doing things differently. Best of Luck Cuill.

first things first...am i the only person who wants to keep saying "kweel" and not "cool"??

in terms of the site itself, i am sort of liking the design / layout of the search, although like agotterer said, it still needs a lot of work -- from what i can tell, popular search terms have pretty darn good results but unpopular search terms return sub-par results (see his example haha)... in terms of a threat to google?? id have to say its not, at least in this stage of the game, but only time will tell i suppose...

The team is impressive, they've been involved in AltaVista, Google, and Ebay. And Anna seems pretty confident that Google isn't going to change, although one of the lead Google engineers came to speak at our school last semester and said that they change the algorithms hundreds of times each year. He said it's not uncommon that they tweak it several times a week. Maybe it's not going through drastic changes, and maybe it doesn't look like a Yahoo portal, but they do continue to improve their algorithms.

It's really cool to see someone branching off and trying something like search, and I actually support Patterson's efforts. But there is a lot more to Google than just their search algorithm technology. They're running with the momentum of the brand, not to mention the massive collection of apps that people loyally follow.

Can you build a better search algorithm than Google that runs faster, with fewer computers, and containing more indeces? Possibly. I happen to know a lot of people who can cook a better cheeseburger than McDonald's, but that doesn't really get you "all the way." Google has criticisms, such as the search logging, which could be used for leverage and could make for easy user acquisitions. However, when you have HappyCamper1 who is (UsingGoogleSearch && UsingGmail), they'll most likely need a better reason to switch search engines because they're already using an e-mail service by the same company.

The important thing is that she gets into this space and attacks it vigorously. No king rules forever. May the best win.

Tweaking the coefficients is not the same thing as changing the algorithm. Google is no more immune from the Innovator's Dilemma than Microsoft is/was.

They don't have to beat Google, they just have to be good enough to get picked up by Microsoft.

I personally found the search results page a tad confusing. Having the results in a grid layout, my first impression was that they were all ads!

.. also, how do I know which result is #1 ? Top left? Middle?

why does it matter which result is #1 if you can find what you are looking for on the first page in the first row or first column?

Well one of the things about Google that I really like is that they really make a lot of effort to make sure that the #1 result is the page you actually want. It's part of the whole 'Feeling Lucky' thing. I'd rather not have to read through the entire search results page to figure out which one is the most relevant.. that's the search engine's job!

In other words, don't make me think. :)

How much do you think: input the right keywords, get results in the top 1,2,3, has to do with the feedback system of using google over the past few years?

I agree on the don't make me think part.

You could be right, perhaps I've been using Google so much that I've become accustomed to it.

But I still think that all search engines should make a lot of effort to ensure that the #1 result is the best result.

I'd rather spend less time looking through search results if possible.

because you have to look somewhere first.

#1 is whatever is the closest to the upper left corner.

I assumed #1 was top left, but which is #2? To the right or down? My guess is that you read down in columns.

The pictures are misleading. For example, #2 for my name is a page about another guy with the same name, but with a picture of me that's not on that page. Very wrong.

I was excited for a second to see that it thought I had an Erdos number of 2, but it was for someone else with the same last name.

I actually liked the grid view, it fits a lot more results on one screen so I can scan them quickly. They should make it auto-fit the screen width.

"i am sort of liking the design / layout of the search"

Did it strike anyone else that this is one of the ideas that Y-combinator said it would like to fund? (beating Google at design in the search space)

Looks as though as it's closed temporarily: "We’ll be back soon...

Due to overwhelming interest, our Cuil servers are running a bit hot right now. The search engine is momentarily unavailable as we add more capacity.

Thanks for your patience."

Now it's not responding at all. Don't sell your Google stock just yet.

They did just launch. I'm sure they still have a few bugs to work out.

"Release early, release often." Rarely balanced properly for potentially high-demand, high-load applications, even if just in spikes.

Searching Cuil for 'cuil' results in a lot of hits, none being their site. Searching Google for 'cuil' has them up at number one, plus news results from today's PR bash. That's a big ball of fail.

My search for 'cuil' on Cuil was even worse, leading to zero results, while a search for 'Cuil' yields lots of results (but, as you noted, not cuil.com itself). So it looks like Cuil is case-sensitive as well as incomplete.

Searching for 'cuill' (two 'l's) works - which is what they apparently changed their name from, seemingly at the last minute.

It seems to be putting incorrect pictures with names. I did a simple search of "Hacker News" and Sam Odio came up on the first search page. I thought, cool, Cuil managed to put his picture right next to his Hacker News profile. But then I clicked through to his blog and unless he's really changed over time, I think they got the picture wrong :)

Take a look: http://billkatz-test.appspot.com/static/images/cuil.png

I did a search for "Terror Island" and although the first link to the webcomic http://terrorisland.net had its title bar correct, the links to individual strips had unrelated pictures of boats.

I laughed. My name came up with some of my online profiles, but with a photo of some British boy band singer....

So how are they planning to monetize this? AdSense?

So far they're vague on this. Ironically Cuil was useless, http://www.cuil.com/search?q=cuil%20money&sl=long http://www.google.com/search?hl=en&q=cuil+money&btnG...

but with google I found this:

"Eventually, Cuil plans to make money by running ads alongside search results, she said, but provided no further details."

The design is nice and the layout/features seem promising, though. Pleasant to use when it works.

Whoever does their PR earned their check today.

Looks good, and has a nice feel.

http://www.cuil.com/search?q=python%20generators&sl=long

But not so good here,

http://www.cuil.com/search?q=python%20generator%20examples&#...

Or here:

http://flownet.com/ron/cuil.gif

(They seem to have fixed this.)

Interesting. I ran a few keywords through and ...

with http://www.cuil.com/search?q=hacker+news we get an url to this site, good.

with http://www.cuil.com/search?q=hackernews we get a link to pretty what I've put up: eg: http://www.flickr.com/photos/bootload/2710119622/

with http://www.cuil.com/search?q=cuil you get no mention of the site itself.

with http://www.cuil.com/search?q=spock you get a peek of how they cuil are structuring more complex searches using drop down menus - bad because it's slower to read & requires a conscious effort to do so.

"Rather than trying to mimic Google's method of ranking the quantity and quality of links to Web sites, Patterson says Cuill's technology drills into the actual content of a page."

Fancy presentation can't save you from the fate of altavista. What are they doing to solve the trust problem?

This has some serious work still...

A search for my name put me in a porno context with an image of a naked girl. The name of the video was "dirty movie", and the html title tag was "dirty movie on vimeo". This was not a pornographic video. Cuil matched a thumbnail image from movieon.com with a man and a woman having sex. This is something I dont want my name associated with.

Second, I did a search, then hit search again on the same keyword and it brought back 0 results. Took about five more clicks to get it to bring up results again. Another search said it found 8,000 results, but only gave me 3 pages worth of browsing. Last but not least the about page (http://www.cuil.com/info/) is not found.

http://www.cuil.com/info/

I can see the info page.. Since lot of people are covering them, they are seeing a spike in traffic and maybe not able to handle it.

I'd be interested to know about some of the cooler pages that are indexed by Cuill, but not Google.

http://www.cuil.com/search?q=paul+graham

Somehow they managed to get a monkey and a basketball scene in the images for a paul graham search. That doesn't seem right to me...

I think that this is an interesting approach. I like that they show more info about each result, but I also have more difficulty scanning through the results. I'm not sure if that's really a net win.

They're new, so I don't want to dismiss them, but if their selling point is their larger-than-Google index, I'm not buying. On my sample query, I got a web proxy in both the first and second pages of results. It's easy to have a large index if you are showing the same thing in triplicate. Google does a great job of narrowing their large index to what the user needs to know. Nothing wrong with some healthy competition, but Google was right to stop (publicly) counting pages.

My initial reaction to Cuil was positive, but that's cause I kinda liked the interface and the first few searches I tried did really well. Lesson learned: Try at least 20, cause some were just horrendous.

That said, there is some promise, and the "related information" tab, while currently only semi-useful does show some promise. The challenge is that people are not going to have the patience for Cuil to be "at least as good as Google". That just doesn't work, and shouldn't

I went to post cuil's comment policy on Edward Tufte's bulletin board thread "Not spying on users should be a feature: The keep-it-and-lose-it hypothesis". Searching ET's site in situ is painful, but Google is pretty good at finding Ask ET threads, so I thought I'd try cuil. Cuil didn't find the thread at all with "Edward Tufte Not spying feature", which has got to be unique.

Also, didn't the google blog recently mention they had passed the trillion mark?

I cant get the result for my newly registered domain name on cuil whereas I can see the result on google. Also no blogspot/googlepages result for your query.

I found the result page lot different compared to google UI. Its paradigm shift and difficult to comprehend. Our brain and eyes are used to google search result look and feel.

It doesn't work. Almost none of my searches worked. I think it gives too much importance to title text.

Cuil seems to be a company of hackers. Just have a look at the founders page: Anna Patterson? wrote 2 search engines (Xift and Recall.archive.org) before the GOOG. Louis Monier? major contributor to at least one search engine (Alta Vista), and also said to have helped eBay.

Right now? Vastly flawed rankings-wise.

http://www.cuil.com/search?q=gstepl

Every response is copying the message from a post I wrote on an inactive blog. More active results don't appear whatsoever.

hmm slightly weird. I searched for a term and didnt get any results. I searched for the same term 10 minutes later and i got results? Anyway, some stuff is not returning any results : "nvidia gtx 280" didnt return any result :(

Would cool.com have been better? :)

http://web.archive.org/web/20061004185947/paulgraham.infogam...

I like how the results layout. They need to work on finding more relevant thumbnails. The thumbnail shown for my own website is completely unrelated to me.

I think they should have single column search results.

how is it pronounced? their name is not memorable.

According to the article its "cool". I agree that it is not a memorable name.

sounds more like kewl

With $33m in venture they could have spared a few Gs for a better domain name.

It's just as memorable as google was on it's first day. A little branding and constantly improving results could make it a player. But, only time will tell. I'm cheering for them.

I think google is more 'spellable' than cuil... or is it ciul or coul or ceul. Hell, the title for the post spells it with 2 'l's

That is pretty ironic considering that google is actually a misspelling -- the correct spelling is "googol".

Of course. But googol isn't the 'expected/intuitive' spelling.

Any product that you have to explain to the user how to pronounce will fail.

I looked at the word and my first thought was "how do you say that" - you don't win people over by making them feel stupid.

Don't overgeneralize. Consider "Audi." While obvious to a German speaker, an English speaker who has never heard the word before will wonder if it should be "oh-DI," "OH-di," "OW-di," or "ow-DI?"

I think Cuil is a decent name, far better than Srchr or some other Web 2.0 contraction. If it overcomes its growing pains and solves all the hard problems involved in search, it'll do well.

Audi started in germany then went global. Is cuil starting in Ireland?

Imagine the way a teenage daughter of yuppies would say the word "cool". Kind of like: [KYOO ill]. That helped me remember it, anyway.

If there are so many people bashing cuil they must be doing something right (privacy features, UI) we just don't know all of it yet!

Could be that we're just jealous of how thoroughly the press covered their adequate-but-not-earth-shattering startup. ;)

One really annoying thing is that page up / page down / arrow keys don't work (sometimes). Maddening.

Does anybody know what happened to that second el? They used to be Cuill for the longest...

something tells me a search engine that mimics the /search?q= url of Google isn't so much innovative.

Google results structure is a win. White background is a win.

I think they are solving the wrong problem.

give it some time folks. google wasnt better than ask.com when it launched.

Yes it was. Ask Jeeves was a stupid "natural language" thing back then -- basically a giant, hand edited, FAQ.

If they add ssl I am sold!

Reply on news.ycombinator.com