In a world filled with Google alternatives, this search tool is different even from them: it's powered by humans. Instead of a server farm that crawls through the entire known Web so it can automatically match Web pages to the queries you type, Mahalo's search results are created by humans, in anticipation of the queries its users will type in.
How can this possibly work? Because, Calacanis says, the top 10,000 search terms account for 24 percent of all searches. If you can create great results for the top results, users will learn to appreciate the difference between machine search results--which are often thrown off by spam and poor-quality links--and human-powered search pages, lovingly created by caring search editors. For the obscure "long tail" queries that make up the 76 percent of search terms, Mahalo will serve up Google results.
In the demo I got last night, in advance of Calacanis hitting the D5 stage today, he showed me a few results that were demonstrably better than what Google would return, both in content and presentation. Searching for "Paris Hotel," for example, gave a list of great links, clearly chosen by someone who knows the difference between a link farm and a real travel site. Also, the links are categorized in the way a human would set them up: by general price category. A search on "Corvette," had similarly good links, as well as RSS feeds from appropriate car fan blogs, a stats box showing information about the current Corvette model year, a list of links to cars that Corvette buyers might also be looking at, and other sections of relevant links and info.
Each result gets its own forum page, so users can chime in if links are missing or bad, and users can also suggest pages for inclusion. All suggestions are run by Mahalo editors, who can decide if they want to put a suggestion in their main page, relegate to a sidebar, or ignore it.
It looks like a very useful service. But at first glance, it also looks like an insane business. Conventional wisdom is that a search engine can't do much good if it's not a machine. And clearly there's no way Mahalo's staff can index the entire Web. But that is not the point. Rather, Calacanis wants his editors to apply their domain knowledge and research skills toward creating resource pages for the topics that are known to be popular (people who want to compete with Mahalo can also buy the same lists of popular search terms, or just keep an eye on Google Hot Trends).
Furthermore, About.com shows (or once showed), that a guide to the Web can work and be successful. Mahalo, in fact, is really not a search engine. Rather, it's a professionally-produced wiki -- that users access via a search query.
The company has 40 fulltime editors and launches today in alpha with about 4,000 results. It will hit beta later this year with 10,000 results. Calacanis hopes that simple editorial excellence will be enough to win over a large user base. As he says, since Mahalo falls back to displaying Google results when it doesn't have something on a topic, "the Google user has nothing to lose by using us."
This one is all about the content. Hit the site too often with what you think is a common term and come up empty, and you'll stop coming back. But if it works when you need it, and if the team can keep those 10,000 pages up to date, well then, great. It could be highly useful, but it will take a year or two for the company to stock its inventory with the content it needs.
Also, if you live near Santa Monica and are looking for a job surfing the Web, I think Mahalo is hiring.