The hype cycle now lasts less than a day. Take yesterday’s over-hyped launch of stealth search startup Cuil, which was quickly followed by a backlash when everyone realized that it was selling a bill of goods. This was entirely the company’s own fault. It pre-briefed every blogger and tech journalist on the planet, but didn’t allow anyone to actually test the search engine before the launch.

The company’s founders have a good pedigree, and have developed a unique way to index the Web cheaply and at massive scale. But creating a big index is only half the battle. A good search engine has to bring back the best results from that haystack as well. Here Cuil falls short, as we pointed out an hour after the site launched and we could actually check it out.

The story quickly turned from Google-killer to Google’s lunch (make that an amuse bouche). The results Cuil returns aren’t particularly great, and sometimes completely off the mark. For instance, a search for “Cuil” doesn’t even bring up a link to itself on the first page of results. (See screen shot at end of post).

And when Cuil tries to pair images with sets of search results, it often chooses seemingly random images to accompany a set of results. For instance, “Wordpress” is associated with what looks like a TV newscast team and “TechCrunch” is paired with a Gmail logo. And I have no idea who that person is next to the results for our “About” section.

The backlash goes beyond the blogosphere. Friendfeed founder and former Googler Paul Bucheit started this thread of deservedly harsh comments on FriendFeed:

Paul Buchheit posted a message
“Maybe Cuil isn’t supposed to be good. They must know that the results are bad, but they launched anyway. Maybe they aren’t trying to build a full search engine, but just want to demo their crawling+indexing technology with the expectation that someone will buy the company and plug in better ranking.”
. . .
—Where “someone” = MSFT, who has already shown that they are willing to pay a lot for non-functional search engines. - Paul Buchheit
—That is what I was thinking, too. – Robert Scoble
—You’d think they’d work toward getting something relevant if you search on “cuil launch” – Michael Markman
—It’s the only thing that makes sense when you look at how bad the results are. You don’t roll out something like unless you are just showing off the interface. – Kevin Bondelli
—Would also explain the weak branding. Who cares, if it’s just going to get plugged into something else? – Chris Baskind
—Not very good results at the moment I will say – shinchi via twhirl
—It must be so hard to launch something in a realm where there isn’t much tolerance/patience for incremental improvement. The bar in this space is high and consumers are very picky. Look at Yahoo. Their search is actually pretty damn good. However, they keep losing share. – Sacca
—agree with Sacca – also, I thought the people behind Cuil had already sold some search technology to Google and hence they wanted to try it out on their own this time – so if anything they’d have more intent than others not to flip. – Adam Kazwell
. . .
—Google Killer? I do not even need to look to know the answer – Mike Reynolds
. . .
—It’s embarrassing IMO. The results are thin and the images completely wrong. Not good is one thing … Cuil seems a cut below not good right now. – AJ Kohn
. . .
—IMHO there have to be better ways to show off your very own crawling+indexing technologies than opening up to the public and getting thrashed for not delivering what people expect when you dub your service search engine. I have a hard time believing that the cuil/cuill guys did not know what they were doing though … Were they pushed to release by financiers? Were they in dire need of usage data? – Mustafa K. Isik

If you are going to feed the hype cycle, you’d better be able to deliver. Because it only lasts about 20 seconds if you don’t.

(Photo by Steve Jurvetson).

The new Cuil search engine apparently got a bit more traffic than the team anticipated immediately after launch a couple of hours ago. Everyone is trying it out to decide for themselves how disruptive it may be to the old guard search guys. For now, you’ll have to wait, a message on the site says “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.”

Launching a startup ain’t easy. And flatlining right after your launch is more of a rite of passage than an embarrassment. Here’s hoping Cuil is cool again by morning.

Update: Yay! It’s back.

Search engine Cuil launched earlier this evening, claiming a bigger index size (120 billion web pages) than Google or any other search engine. The pedigree of the founders and execs, which includes three ex senior Googlers, means the service will be compared to Google from day one. And the way they will be compared is index size and, more importantly, relevance/ranking of results.

We’ve been testing the engine for the last hour. Based on our test queries Cuil is an excellent search engine, particularly since it is all of an hour old. But it doesn’t appear to have the depth of results that Google has, despite their claims. And the results are not nearly as relevant.

A search for Dog returns 280 million results on Cuil and 498 million on Google. Judging relevance of results is subjective, but Google returns Wikipedia as the first result, then dog.com. Cuil returns Dog.com, wikipedia isn’t listed on the first page of results. Both are meaningful results, but Google is better.

More searches, Cuil v. Google: Apple (83 m v. 571 million) – neither mention the fruit. France (102 m v. 1.5 billion) – Cuil’s category refinement makes their results better for this query. Stonehenge (800k v. 8.5 million). Silicon Valley (3.2 m v. 24 m). Techcrunch (600k v. 6.5 m).

It seems pretty clear that Google’s index of web pages is significantly larger than Cuil’s unless we’re randomly choosing the wrong queries. Based on the queries above, Google is averaging nearly 10x the number of results of Cuil.

And Cuil’s ranking isn’t as good as Google’s based on the pure results returned from both queries. Where Cuil excels is with the related categories, which return results that are extremely relevant. With Google, we’ve all gotten used to trying a slightly different search to get the refined results we need. Cuil does a good job of guessing what we’ll want next and presents that in the top right widget. That means Cuil saves time for more research based queries.

And I want to reemphasize that Cuil is only an hour old at this point, Google has had a decade to perfect their search engine.

Menlo Park based Cuil will launch later this evening with an index of 120 billion web pages, making them arguably the most comprehensive search engine on the web (Google doesn’t disclose the size of their index, although they claim to know about a trillion unique web pages) (Update: see our very early testing here). They’ve also dropped one of the “l’s” from their name – previously the company was “Cuill.” Either way, it’s pronounced “cool.”

The super-stealth search project was founded by highly respected search experts. Husband and wife team Tom Costello (CEO) and Anna Patterson (VP Engineering) were joined by Russell Power. Patterson and Power are also ex-Google employees, and the company has been the subject of intense speculation over the last couple of years.

Much of the secret sauce of Cuil is in the way they index the web and handle actual queries by users. Both are costly to scale, and Cuil claims to have found a way to massively reduce those costs. That allows them to run the search engine a lot cheaper, even at Google-scale should it ever reach that point. By some estimates, Google spends a billion dollars a year to run the back end infrastructure of it’s search business.

Cuil also claims to have better search results than Google and others based on how they index websites. They do not simply catalog keywords on a site and then rank the site based on its importance. They also work to understand how words are related (France – cheese – wine, for example), to return more relevant results to users. This is a semantic approach to search, but very different from Powerset’s natural language approach. Powerset uses artificial intelligence to try to understand what sentences on a website actually mean. Cuil, by comparison, simply tries to properly categorize and file a web page, even if the category name doesn’t appear on the site.

That means users search the same way they always have, but Cuil will try to return better results via refinements in a “explore by category” module to the right of results. A search for dogs, for example, will return category results for “water dogs,” “crossbreed,” “cocker spaniel,” etc. Some of these related terms do not include the term “dog.”

Cuil is experimenting with a new type of search interface as well. Results are shown in three columns and contain an image and more summary text than existing search engines. In addition to refinement by category, Cuil will recommend related searches via tabs across the top of search results. A search for New York, for example, also has tabbed results for recommended refinements like New York Times, New York City, New York Yankees, etc.:

Cuil also says that they will put user privacy at the top of their business objectives. User IP addresses are not recorded to their servers, they say, and cookies are not used to associate a computer with queries. The data is simply dumped as it is created. That means user data cannot be turned over to others, whether its via blind stupidity or lawsuits.

Cuil has raised $33 million over two rounds of financing from Greylock, Madrone Capital Partners and Tugboat Ventures.