Kagi is a paid alternative to ad-supported search engines like Google and DuckDuckGo. It has recently revised its pricing model, reducing the cost for a plan with unmetered searches from $25 per month to $10.

Kagi boasts the following (and more) features:

  • Blocking or boosting specific domains in your search results
  • “Lenses”, which are individual setting profiles (e.g. region locks, domain whitelists) that can be applied to search queries
  • All of the Bangs that DuckDuckGo has (e.g. type “!yt” in front of your query to immediately search on youtube.com)
  • Universal Summarizer, which works with any website, PDF document, YouTube video and more

This blog post goes into full details about Kagi’s capabilities.

  • That’s a curious project and I hope they succeed. But I have to wonder. On their “Why pay for search engines” page, they state the following:

    Our proposed price is dictated by the fact that search itself has a non-zero cost. In fact, it costs us about $1 to process 80 searches (wherever in the world you search from). So a user searching 8 times a day would perform about 240 searches a month, costing us $3 in search cost. But an average Kagi user is actually searching about 30 times a day. At USD $10/month, the price does not even cover our cost for average use.

    So, will they dial the price back up or do they currently just hope that most people pay for the “unlimited searches per month” plan but use it less than an average user would?

    •  Dave   ( @Dave@lemmy.nz ) 
      link
      fedilink
      179 months ago

      They probably haven’t updated the page. This blog post says:

      With new search sources proving more cost-efficient, the improved efficiency of our infrastructure, and the broader market embracing Kagi, we can again offer an unlimited experience to a broader group of users.

      So it sounds like they have made lots of efficiencies to make it cheaper per search. I’m sure more subscribers helps as well.

      But I’m really curious about the “new search sources” part. Where do they source their searches from?

    • They probably have a lot of potential infrastructure savings. $1/80 searches is an absolutely astronomical cost.

      I’m imagining there are quite a few gains they can get by way of optimization, different technologies, and optimizing hot paths to bring that number down.

      It really depends how they built this thing. For instance, if they built this on the AWS ecosystem, using more than straight compute/K8, their costs are going to be an actual order of magnitude higher than if they didn’t.