Mahalo

» see mahalo.com as it lived, on the wayback machine

~ the goodbye ~

mahalo's pitch was that humans could outperform google's algorithm at building search results. google disagreed within four years. jason calacanis launched the site on may 30, 2007 as a "human-powered search engine," with paid editors hand-building search result pages for popular queries. the listing on each page was the seven sites a real person had vetted, instead of whatever google's pagerank was returning that week. mahalo paid editors at $10-15 per page and at peak had hundreds of them on the books. for a year or two in 2008-2009 it was the loudest argument anybody made for the human-curation thesis.

google's panda update on february 23, 2011 sent mahalo's traffic off a cliff. the algorithm was designed specifically to demote thin content farms, and mahalo's wage-piecework "how-to" articles were exactly the kind of content panda was looking for. by mid-2011 the company laid off most of its writing staff. by january 2014 calacanis announced "server sunset," meaning mahalo would keep its servers up but receive no further engineering investment. he pivoted to a mobile news-curation app called inside, which itself faded a few years later.

the mahalo idea was older than mahalo. yahoo directory had done human-curated web links since 1994. dmoz did it from 1998 until 2017 with volunteers. what calacanis added was the willingness to pay piecework wages to fill out long-tail query coverage. that was the part the google panda update treated as junk. it turned out the algorithm did not need a human in the loop nearly as much as the 2007 pitch deck assumed.

i was a computer engineering student in 2007, paying a little attention to the english-language web 2.0 conference circuit through techcrunch. mahalo was one of those names that came up at every panel for two years and then never again. by the time it sunset i had to look the date up. that is the standard ending for the calacanis genre of startup.

~ the chronicle ~

~ the breakout ~

calacanis raised $20 million from sequoia and others in 2007, on the back of his prior weblogs inc sale to aol. the launch coverage was enormous. techcrunch and every tech blog of the era covered each new feature like a state visit. mahalo answers, the q&a product launched in december 2008, paid users in real money for good answers and was briefly the most credible-looking yahoo answers competitor in the english web. that one peaked early and faded by 2010.

~ panda ~

when google rolled out panda on february 23, 2011, mahalo's traffic dropped between 40 and 70 percent depending on which post-mortem you read. the company had been building "how-to" pages at piecework rates specifically to capture long-tail google searches. panda treated those pages as the exact kind of low-quality affiliate content the update was designed to suppress. there was no recovering from being on the wrong side of that taxonomy. by june mahalo had laid off ten percent of its staff and pivoted to video tutorials, which did not go any better.

~ what got buried with it ~

the mahalo era was when the english tech press still half-believed humans could beat algorithms at building the index of the web. calacanis was the loudest version of that argument. when his site got panda-flattened the argument went with it. the "human curation will save us from algorithmic slop" thesis is back in 2026 in the context of llm hallucinations, but it is now an aesthetic preference rather than a business model. mahalo is what tried, and what failed, the previous time around.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »