Now Cuil is back with a new project I learned about in a GigaOM post by Matthew Ingram: Cpedia, an algorithmic encyclopedia with more than 384 million “automated articles.” Cuil founder and CEO Tom Costello explains:

Cpedia returns an automatically written article in response to a query, rather than a list of hits. It can be very compelling; it is especially good at surfacing facts that I didn’t know before. At other times it is weird — it does reflect the web after all.

[section explaining that Cpedia censors offensive matter snipped]

I find Cpedia best on topics that I thought I knew about. I find out things I should have known but didn’t. I’ve noticed productivity has slowed in the company since we have had it up for internal testing, as people ask each other about stranger and stranger trivia, or exclaim, “I didn’t know your middle name was Hector?”

Cpedia is very different from a traditional search engine, and not at all like Wikipedia, but that is its strength; it is something new and different. I hope you like it. I certainly do.

Different? Weird? I’ll say. So far, the best articles I’ve checked out consist of brief excerpts from other sources–often only vaguely related to the alleged topic–that have been assembled in seemingly random order into a jumble of paragraphs that are robbed of their context. The worst entries? Even the sentences in them are unintelligible.

Here are excerpts from a couple of typical entries:

The entry on Cuil itself is rife with gibberish (“Clickbooth Cuil but not avail due to flooding traffics…”)

Here’s a section from the entry on…me:

Cuil’s executive team boasts former Google and IBM search scientists, and multiple PhDs from Stanford. And maybe the company has an idiosyncratic sense of humor. Given that Cpedia launched on April 8th, I’m not discounting the possibility that it’s all an elaborate practical joke. It certainly tops most of the stuff here…

It’s not like nobody’s giving it all they’ve got. Cuil is merely the most recent startup to be positioned as a possible Googleslayer: Others have included Powerset (recently snapped up by Microsoft), Wisenut (which is no longer with us), and Wikia. And every time Yahoo or Microsoft or Ask launches some feature which will supposedly prove an irresistable lure to Google fans, Google’s share of all searches only trends upward.

In short, nobody’s really even managed to give Google a flesh wound. As with Freddy Krueger, tryng to kill it seems to do more harm than good. But why?

1. The very audacity of the goal makes it nearly impossible. As Cuil shows, you can have a sterling management team, millions of dollars, and theoretically powerful technology and still come up with a mediocre search engine. Coming up one with that’s clearly better than Google is as ambitious as a goal can get. It’s like deciding to go to the moon in 1961–and when JFK declared that we’d do that, he sensibly gave the country eight and a half years to get the job done.

If I were starting a search engine, I’d cheerfully declare that my only desire was create something decent enough to get on the map–even if I had hopes of eventually crushing Google and thereby becoming a trillionaire.

2. You can’t do it through adding features. Google’s minimalism is its greatest asset. You don’t need to study up on how to use it; there are few hidden tools or advanced functions; on the surface, it’s amazingly close to the search engine it was a decade ago. The company has shown amazing discipline in the way it’s shunned clutter and complexity.

(At PC World, we used to discuss doing a cover story on Google Tips, but we kept deciding that the thing was so basically simple to use that we couldn’t come up with enough advice on how to get more out of it.

Ask.com is instructive here: The search engine formerly known as Ask Jeeves provides search suggestions as you type. It lets you narrow or expand your search terms. It melds disparate types of content like news, images, business information, and Wikipedia listings together on one page in a way that’s more elaborate than Google’s Universal Search. A feature called AskEraser lets you delete your search queries from Ask’s servers. You can skin the search engine with polka dots or a photo of an old Fiat in Rome.

It’s all pretty cool, and some of it is superior, in theory at least, to Google’s way of doing things. And despite it, Comscore says that Ask’s share of the search market has dropped from 5.0 percent to 4.3 percent since May of last year. (Meanwhile, it says Google’s share is up from 50.7 percent to 61.5 percent.)

It leaves you thinking that the one clear way to beat Google would be by releasing something even sparer and simpler that somehow delivered better results. I’m not saying that’s impossible, but it does seem unimaginably difficult.

3. It’s hard to beat Google by being more Googley than Google, but being different may mean being worse. Sorry to keep beating up on Cuil, but it departs strikingly from the Google way of doing things with its grid-like results page. It gives Cuil a certain amount of personality–but makes it harder to use than Google’s simpler, more straightforward approach.

4. Google isn’t one search engine–it’s a lot of them. It’s great at local search. It’s got a decent blog search engine. Google News is far from perfect, but still mighty handy. It owns YouTube. And it’s tying everything together by weaving results from all its search engines into one results page.

A Microsoft or a Yahoo or even an Ask can at least aspire to provide results of all these sorts, even if none of the major search engines have provided ones that are enough better than Google’s to put a dent in it. But even a very well-funded startup can’t launch an array of search products at once. Even if Cuil had nailed Web search, it would solved only one piece of the Google-slaying puzzle.

5. The Google money machine is a virtuous circle. Nobody else is remotely as good as Google at making search pay off through context-sensitive advertising. Which means that nobody else has as much dough to throw at making the search experience better. The fact that Google is the dominant search engine means that advertisers are more likely to give it money; that money helps make Google even more dominant; the more dominant it gets, the more advertisers show up.

It’s true that money alone can’t buy success–if it could, this story would be about how to beat Microsoft’s Windows Live Search–but it sure hasn’t hurt in Google’s case.

This morning, it’s back up and running. But despite claiming the planet’s largest index, it still has some sort of glitch that results in it sometimes failing to find any results at all:

In other Cuil news, my friend Dan Tynan is rightly wondering what Cuil’s founders were thinking when they gave their site that name. They say it’s an old Irish word for knowledge, and that it’s pronounced “cool.”

It seems inevitable that half of Cuil’s users will have no idea how to pronounce it. And if any of the the other half recommend it to a friend in person, that friend will have no idea how to spell it unless it’s spelled out–which is kind of a singular achievement for a word with only four letters in it.

It gets weirder: Until recently, Cuil was going to be called Cuill. They say that they changed the name to simplify the spelling…

To be named Cuill and change the spelling to Cuil based on that rational is sorta like living near the San Andreas Fault and deciding to move to safer territory–and buying a house that’s fifty yards to the south.

Safe prediction: It Cuil turns out to be good enough to gain traction, it will be called something else someday. Maybe soon…

After spending a bit of time playing with Cuil, though, I’m more puzzled than impressed by the results. It would appear that the site is suffering some technical glitches tonight: In some cases, it’s told me it found zero results for a search, then has returned lots of them when I tried again. In fact, that’s happened often enough that I’d be cautious about judging any results that Cuil provides tonight:

I’ve done a bunch of other searches in hopes of finding an instance in which Cuil clearly beats Google. No luck so far. When I search for George Washington, the first result relates to George Washington University, not the Father of the Nation. The second result is Wikipedia’s entry on the great man, but the text excerpt is a snippet from the bibliography at the end of the article, so that’s not clear. The third result is a page about the George Washington Carver museum in Austin, Texas. I can’t imagine anyone arguing that that outdoes Google’s results. In fact, placing a result relating to George Washington Carver so high is evidence that Cuil’s understanding of my search was shaky; it’s unlikely that anyone looking for information on George Washinton Carver would fail to include the “Carver” in the search.

When I tried to pull up info on a local Thai resaurant with the search “osha restaurant san francisco,” all the results on Cuil’s first page were from Citysearch, and they included dupes. Google gave me the restaurant’s official site, the addresses of all its locations, reviews from Yelp, and more–including a link to Citysearch.

Cuil’s first result for “america” is AOL; its first result for “capitol” is for the Nebraska state capitol; its first result for “hamburger” is a site selling a book of restaurant reviews for airplane pilots.

Oh, and its first result for “iphone” is for a phone from…Cisco. (McCracken’s Fifth Law of Search Relevance states that zero percent of people who search for “iphone” are seeking information on Cisco VoIP products.)

Other Cuil results are also odd: When I search for “cuil search engine,” I get one result–which isn’t about Cuil. It’s true that the site was formally known as Cuill, with two Ls at the end. But when I do the same search on Google, I get plenty of relevant results.

There’s some evidence that Cuil’s index, whatever its size, isn’t super-current. When I searched for “dark knight,” I rightly got a link to the official site as my first result. But Cuil’s text snipped for it referred to it as an upcoming film, and most of the other links on the page were similarly stale. Google’s results were much more current.

Cuil displays its results in a manner which is decidedly un-Google-esque: You get a gridlike page with ten or eleven results laid out in rows and columns. The results give you a larger-than-usual excerpt from the pages they link to, but I’d love to know Cuil’s thinking behind using the grid rather than a traditional top-to-bottom list: For me, at least, the Cuil approach is dramatically harder to scan, since I can’t just swoop my eyes down the page.

The search engine does have some features designed to help searchers. As you type, it begins to list searches that include the characters you’ve entered: For instance, when I typed “honda c” it presented me with this

You also get tabs that provide access to related searches:

And an “Explore by Category” box that provides links to further related results:

None of this is strikingly new, though. Here’s Yahoo doing something quite similar, for instance:

Then there’s Cuil’s claim of having the Web’s largest index. Over at TechCrunch, Michael Arrington did some searches for which Google returned more results than Cuil did. But judging a search engine by the size of its index is a basically iffy proposition. The search engine with the biggest index may have done a poorer job of eliminating dupes like the one I found in my Thai restaurant search above. And getting millions of results for a search doesn’t help you a bit when even the most dedicated of searchers don’t look at more than the first few pages of results. (Millions more pages in the index is meaningful only if some of those extra pages are so relevant that they show up near the top of some results; otherwise, they’re just dross.) Which helps explain why when Google announced last week that its engine knows of a trillion URLs on the Web, the company also said that many of them are irrelevant and therefore not in the index.

None of this is to suggest that Cuil might not turn out to be something good. I cheerfully admit that when I first used Google–back when it was an obscure upstart with a lot of buzz, not unlike Cuil–I was distinctly unimpressed and went back to whatever search engine I favored at the time. (AltaVista? Hotbot? I forget.) It simply took awhile before Google became Google.

In fact, given the pedigree of the folks behind Cuil and its underwhelming performance in most of the searches I tried, I think the technology more or less must be better than the experience I got tonight. So I’ll give Cuil another try soon. For now, though, I’m surprised that this is one new site that doesn’t have a little “Beta” label next to its logo. Actually, what I experienced tonight felt like a shaky alpha, even if there’s powerful technology lurking inside there somewhere…

(Update: I just tried Cuil again and got a “We’ll Be Back Soon” page saying that Cuil usage is outstripping capacity.)