Kagi Search Beat Down

April 17, 2024

This essay is the work of a dumb dinobaby. No smart software required.

People surprise me. It is difficult to craft a search engine. Sure, a recent compsci graduate will tell you, “Piece of cake.” It is not. Even with oodles of open source technology, easily gettable content, and a few valiant individuals who actually want relevant results — search and retrieval are tough to get right. The secret to good search, in my opinion, is to define a domain, preferably a technical field, identify the relevant content, obtain rights, if necessary, and then do the indexing and the other “stuff.”

In my experience, it is a good idea to have either a friend with deep pockets, a US government grant (hello, NSF, said Google decades ago), or a credit card with a hefty credit line. Failing these generally acceptable solutions, one can venture into the land of other people’s money. When that runs out or just does not work, one can become a pay-to-play outfit. We know what that business model delivers. But for a tiny percentage of online users, a subscription service makes perfect sense. The only problem is that selling subscriptions is expensive, and there is the problem of churn. Lose a customer and spend quite a bit of money replacing that individual. Lose big customers spend oodles and oodles of money replacing that big spender.

I read “Do Not Use Kagi.” This, in turn, directed me to “Why I Lost Faith in Kagi.” Okay, what’s up with the Kagi booing? The “Lost Faith” article runs about 4,000 words. The key passage for me is:

Between the absolute blasé attitude towards privacy, the 100% dedication to AI being the future of search, and the completely misguided use of the company’s limited funds, I honestly can’t see Kagi as something I could ever recommend to people.

I looked at Kagi when it first became available, and I wrote a short email to the “Vlad” persona. I am not sure if I followed up. I was curious about how the blend of artificial intelligence and metasearch was going to deal with such issues as:

Deduplication of results
Latency when a complex query in a metasearch system has to wait for a module to do it thing
How the business model was going to work: Expensive subscription, venture funding, collateral sales of the interface to law enforcement, advertising, etc..
Controlling the cost of the pings, pipes, and power for the plumbing
Spam control.

I know from experience that those dabbling in the search game ignore some of my routine questions. The reasons range from “we are smarter than you” to “our approach just handles these issues.”

Thanks, MSFT Copilot. Recognize anyone in the image you created?

I still struggle with the business model of non-ad supported search and retrieval systems. Subscriptions work. Well, they worked out of the gate for ChatGPT, but how many smart search systems do I want to join? Answer: Zero.

Metasearch systems are simply sucker fish on the shark bodies of a Web search operator. Bing is in the metasearch game because it is a fraction of the Googzilla operation. It is doing what it can to boost its user base. Just look at the wonky Edge ads and the rumored miniscule gain the additional of smart search has delivered to Bing traffic. Poor Yandex is relocating and finds itself in a different world from the cheerful environment of Russia.

Web content indexing is expensive, difficult, and tricky.

But why pick on Kagi? Beats me. Why not write about dogpile.com, ask.com, the duck thing, or startpage.com (formerly ixquick.com)? Each embodies a certain subsonic vibe, right?

Maybe it is the AI flavor of Kagi? Maybe it is the amateur hour approach taken with some functions? Maybe it is just a disconnect between an informed user and an entrepreneurial outfit running a mile a minute with a sign that says, “Subscribe”?

I don’t know, but it is interesting when Web search is essentially a massive disappointment that some bright GenX’er has not figured out a solution.

Stephen E Arnold, April 17, 2024

Written by Stephen E. Arnold · Filed Under AI, News, Search

Comments

Got something to say?

Search the site
Subscribe to Beyond Search
Feature archive
News archive

Stephen E. Arnold monitors search, content processing, text mining and related topics from his high-tech nerve center in rural Kentucky. He tries to winnow the goose feathers from the giblets. He works with colleagues worldwide to make this Web log useful to those who want to go "beyond search". Contact him at sa [at] arnoldit.com. His Web site with additional information about search is arnoldit.com.