The press picked them up before they were actually ready to launch. Their search model was to segregate results into different categories and show all the categories at once, in sections of the page. When the traffic spike, the most popular categories went offline, leaving only the long-tail sections active. Most people never saw how the search would have looked when the software was finished and the servers beefed up.
At cuil we were told by mgmt of the launch date and our very talented PR team kicked into gear and generated a ton of buzz via a nationwide press tour.
Despite popular opinion, the ops team kept the site up during the massive traffic surge. A combination of poor mgmt, initial deference to user testing, last minute commits and a nasty indexing bug were the reasons the relevance sucked on day 1.
