I’ve been experimenting with Kagi since March, and subscribed to it, and set it as my default search engine, in May. I thought I’d written up a blog post about that, but I can’t find one. I’ve mentioned Kagi in passing, but I don’t seem to have written a post specifically about Kagi.
I’m on their $5/month “starter” plan, which gives me 300 searches per month. Since Kagi is blocked at work and I can only use it at home, I thought that would likely be enough. Today, I got a notice that I’ve hit my 300 searches and would have to either upgrade or renew my monthly plan early if I wanted more. I like the idea of renewing early; it gives you an out, if you don’t think you’re ready to upgrade but want to reset your allocation and keep searching.
I checked to see where I was in my billing cycle, and my subscription should renew today anyway, so I’m just going to wait and let that happen. It might not renew until end of day, so I guess I’ll just use Bing or DuckDuckGo today. Overall, I’m finding Kagi to be a really good search engine, and going back to any other search engine is now kind of an annoyance.
Kagi also has AI features, though they’re not pushed on you the way they are with Google or Bing. I had some fun this week asking various AI assistants about the existence of a seahorse emoji, after reading this blog post. Kagi’s was the only one that gave a succinct and correct answer:
There is no seahorse emoji in the Unicode Standard. The belief that one exists is a common misconception, often attributed to the Mandela effect.
Most of the others got confused, to various levels. ChatGPT got the most confused and really went wild. Here’s a link to the chat session for that. If I ask it the same question today, it seems to get just a little confused, then gives me the right answer. I assume the seahorse thing went viral enough that the various LLM chat companies have tweaked things now to prevent freakouts.
For work use, I’m mostly limited to Copilot for AI usage. At home, I’ve experimented with a few, including ChatGPT, Claude, and Poe. And I recently got curious about Perplexity and created an account there. There’s an offer now from PayPal where any PayPal or Venmo user can get a free year of Perplexity Pro. I haven’t decided to sign up for that yet, but I thought I should see if Perplexity was any better/different from the other options.
Overall, Perplexity is interesting, but I’m not sure if it’s worth pursuing. They do seem to emphasize integrating search results over having the LLM generate an answer in a vacuum. So I like that. I’m not sure if it’s better than Copilot or Kagi really, though.
I think I’ve decided to focus most of my home AI use, at this point, to Copilot and Kagi Assistant. I’m paying for both, in a sense. My Microsoft 365 subscription gets me higher usage limits for Copilot than if I was a free user. And my paid Kagi subscription includes a certain amount of AI usage. So I want to see if I can focus on using those two tools effectively, vs switching around between the four or five tools I’ve been playing with recently.