Hacker News Reader

Comment by strogonoff | Hacker News Reader

I’m wary of making Kagi a habit since I am positive in medium term they won’t be profitable with their current model. Flat rate won’t cover outliers and abusers, so once they really start losing money on expensive LLM-based features the regular users will get burnt—either I will have to pay more, or the experience will get worse due to cost-cutting measures, or they betray my trust by monetizing user data, accepting more investment which never comes without strings attached, or just selling the business.

The best possible, most honest and sustainable option would be usage-based billing, but it appears that Kagi have decided against that route.

I’m a moderately happy customer and would hands down recommend Kagi, but personally wouldn’t use it by default as the inevitable cost of changing my habit again in future puts me off.

View on HN

On the flipside, I whole-heartedly recommend Kagi because I haven't gotten the sense they're running an unsustainable VC-backed business. As long as Kagi exists and provides such a competent search engine, I'm willing to give them my money monthly. I don't see an alternative if I want something other than Google or Bing.

They've repeatedly adjusted their pricing and usage rate billing based on their needs so that I don't feel there's a real risk from abusers/exploiters. They seem to be pragmatic enough to clamp down on that kind of thing as soon as they see how the costs develop under current circumstances and caused by what users/user habits.

I have that faith in them because so far they've been very transparent in their blog posts about when they've made changes and why.

Just drop them when they fail/become bad. No reason to miss now for some future eventuality which does not affect the present.

Yeah it's a philosophy I've started to adopt. Better to get in when the product is new and there is hype, to enjoy fast-paced improvements from a new competitor that just wants to create a cool product, than to come in later when the hype has died down, the original founders have exited and the enshittification starts.

We all know the hype cycle of programming languages. Hype, disillusion then plateau of productivity. It's not the same with startups: there is only hype, peak, followed a long descent to the mean and death.

The hipsters were right.

Not sure if that letdown will trump whatever minor time savings I have accrued while it was good, though. To make it worse, first make it better, then make it back as before.

It’s the same reason I choose not to make use of insane promotions that ride app space newcomers tend to provide. I could use Bolt DeLuxe (or whatever is the new hotness in town) for a couple months, but the anticipation of how I will soon suffer as it becomes more and more of a ripoff with incompetent drivers and absent customer support, and how I’ll waste willpower on changing my habit again, makes me wonder if it’s really worth it. Instead I could in meantime exercise my taxi hailing skills (in this analogy, practice crafting queries and otherwise working around the limitations of worse search engines).

Heya, I work at Kagi, and I did the math for this. Our unit economics are sound, and we don't plan on subsiding usage with ties-attached money. We have abuse prevention mechanisms, and we regularly review heavy use use-cases so we can either optimize for them or offer alternative workflows.

> they won’t be profitable with their current model. Flat rate won’t cover outliers and abusers

They went from "a few tiers + pay-per-use above that" to "one low tier + flat rate unlimited" not long ago, don't you think they a) thought it through and b) are flexible enough with their pricing and profitability?

Quite the opposite: I know they thought this through (thanks to them being quite transparent in their communication[0]), and that their own conclusion was:

> the $10 per month unmetered search plan was not financially sustainable for us.

This was in March; fast forward a few months and surprise, they offer a $10 per month unmetered search plan. With unmetered LLM usage to boot?

I think it is obvious from their own math that the only way flat rate would be profitable is if they go for growth-based approach, burn whosever investment money for a bit and accumulate enough light users that they end up fully subsidizing the power users. I.e., they are hoping that the tragedy of the commons will not apply to them. Good luck to them, I guess.

(That, or they have the infrastructure to enforce some limits on unlimited plans. Unlikely, as it is not mentioned on the pricing page. If true, this would mean some invisible and inflexible limits I would have to be constantly mindful of running against—which is significantly worse compared to pay as you go.)

[0] https://blog.kagi.com/update-kagi-search-pricing

We did not have any investment money to burn, Kagi was completely bootstrapped until June when we raised a very small amount from our users.

What changed is:

- Price of LLM APIs dropped significantly faster than we anticipated

- Kagi started getting adopted more broadly, which in turn meant more people who do not search as much started using it, lowering our average cost per user

Combination of those two factors made us more comfortable to adapt pricing again and pass all anticipated savings to our members.

Thanks for the reply! My concern is with counting on people who do not search a lot to be a big part of your user base. Flat rate means those people might feel like they are overpaying, and if they leave you will start losing money on people who search a lot.

Being concerned is healthy, there are no worry free businesses.

There are no worry free businesses, but there are more and less sustainable business models and different attitudes to existential risk management.

Kagi at first seemed strategic with metered paid search, but choosing to go for growth and hoping most users will not search a lot feels like a downgrade from that.

> usage-based billing

US corporations doing this never offer proper spending limits per timespan (f.e. Amazon and Microsoft offer this but don't offer real limits - their wording explicitly allows them to go infinitely over and you will just be fucked). If they started doing this I would probably unsubscribe tomorrow (I'm currently a paying user), the culture just isn't there for them to get that right.

They can implement a system where you top up your account with credits, like Skype and many others have.

Fun fact the LLM features are not expensive to them compared to their search product. Per search they pay at least 2ct.

Interesting, I wasn’t aware. I recall there was a post (from Vlad probably?) where they noodled over pricing plans and IIRC they considered low flat rate insufficient, I thought ML would be it but perhaps it’s the search.

Reply on news.ycombinator.com