Comment by MichaelZuo
7 months ago
It really does seem like you’re being a bit too unfocused Vlad.
Delivering high-quality search over the entire Internet, higher quality than Google, is something so complex that even if you were literally the worlds smartest person and all the other Kagi employees were number 2 to number 26, there would still likely be stumbles at least once a year on something.
Because there’s like a million gotchas hidden along the path to just reliably matching Google search quality circa 2010 in the 2024 environment. Let alone delivering a high-quality browser, AI tools, etc., on top.
Serving optimally performing ads to billions of global users is a radically different problem than serving optimal search results and accessory features to a self-selected 50,000 or 500,000 customers.
Hence why I specifically said search quality without mentioning a large userbase...
Did you not see the last part on your end?
Plus, if anything a small userbase makes it more difficult for quality search because the long tail is still effectively infinitely large, relative to the competencies of a single decision maker, but now there is only have one user searching for any random super niche topic maybe once a month, in total.
So they can't even a/b test or rely on customers reporting in on the real situation because it is too sparse.
The point is that search quality is subjective, not objective, and the two companies are each structured to approach it very differently.
In pursuing billions of global users across all demographics and trying to maximally monetize them through ads, Google is pursuing an entirely different measure of "search quality" than Kagi.
Google delivers their version of search quality when a rice farmer in Thailand and financier in the Bay Area both reach for Google when they want to find something online and then get distracted by an ad.
Meanwhile, Kagi gets their version right when they have a profitable base of happy customers. They can make different and more aggressive assumptions about the needs of their users, solicit and digest direct feedback about those assumptions, and optimize a product that delivers superb search quality for their niche.
They're completely different technical problems that only occasionally intersect. Their engineering teams aren't competing with each other.
1 reply →
I'm not convinced it's actually that complicated.
Google search has been bad for a long time. It's clear they serve their customers (advertisers) quite well, but as a user of their search, they're not particularly impressive.
The biggest problem is a problem of scale: being the biggest search provider means Google are targeted by SEO, so it's harder for Google to sift out the AI-generated garbage--Kagi just isn't involved in the arms race that Google is. But as a user that's not my problem; I'm not going to tolerate bad search results out of some sense of "fairness" to a corporation. And Kagi is delivering real user-centered features which are, frankly, obvious, i.e. Google should be embarrassed that they don't let you filter/prioritize domains or search within lenses like Kagi does.
I could see Kagi being able to sidestep maybe 95% of the 'arms race'.
But even the remainder will trip them up every now and then.
If these periods last only a few minutes it probably won't matter but if it lasts a few days or more then it's very likely to impact customer retention.
> Because there’s like a million gotchas hidden along the path to just reliably matching Google search quality
I don’t know about you, but the reason I’m paying for Kagi is because their search results are reliably better than Google.
That may be because all the spam optimizes for Google, but noneless they’ve already done the thing you say they should focus on.
As long as the search quality isn’t compromised it doesn’t matter to me what they do with my money.