Give option to disable AI features & donate money for FOSS instead.

Ophelia

Hey there,

Thank you for the great product!

I am not interested in AI features & the AI credits that I get every month with my "Professional" subscription.
You surely have a mixed calculation, how much AI usage costs per user -- so it would be great, if you could give me the option to disable AI-features & set AI credits to 0, and donate part of this allocated money to free / open source projects instead.

All the best from Berlin & thanks again for your great work,
Ophelia

Give option in settings to disable AI features & wave AI credits that come with the subscription
Kagi then donates part of this allocated money to FOSS projects

Ophelia

I am not happy that my feature request just gets merged into this thread—I hoped my suggestion was a bit more nuanced than just “remove AI from the product”.

In case I don't misunderstand the UI (see attached), I get free AI credits with my subscription, so there is clearly a budget allocated for this.

In the response posts, from a few months ago in this thread, it was argued that Kagi is not paying for LLM credits with your subscription, but this doesn't seem to be true anymore.

Please correct me if I misunderstand the UI—I would hope, that I get a proper reply as a paying customer & not just a “there are 100 other customer requests that sound similar; have fun reading them”.

So, please re-read my initial message & I would hope to get a proper reply to this, since I tried to be constructive.

RoxyRoxyRoxy

Apologies, not sure if I merged or who but you're right this is markedly different. The "cost" you see there isn't something we just pay to AI companies, it's the quota for the fair use policy on your selected plan. If you don't use AI we don't pay for AI.

Kaguy

RoxyRoxyRoxy Sorry if this isn't the right place, but could you please elaborate on what you mean by "if you don't use AI we don't pay for AI"?
I'm asking because surely there's some cost involved with all those AI features you've been adding to the plans, both for developing and for users who use it. Are those credits a way to let the user "pay" for their share of use for those features irrespective of the subscription?

RoxyRoxyRoxy

What AI features do you mean specifically? Quick Answer and Assistant run off that fair use budget. We don't pay any money towards AI providers from that budget unless you use those.

Ophelia

RoxyRoxyRoxy

it's the quota for the fair use policy on your selected plan. If you don't use AI we don't pay for AI.

But it also means that if I were to use it, you would pay for it, right? & since it's under "fair use", I wouldn't pay for it -- which leaves you with money paid out of your pocket.

Correct me if I am wrong. 😄

So I would propose to instead allocate my theoretical usage costs for FOSS. ☺️

Ophelia

this post is left empty, since it didn't use the reply feature correctly & I can't delete it 🙈

Luis

Ophelia there's a misconception of our unit economics. That AI "budget" is to enforce our fair-use policy.

Our search subscriptions (non-Ultimate) don't include extra AI budget. Searches are orders of magnitude more expensive than tokens. Our hypothesis is that AI features like Quick Answer or access to Kagi Assistant have utility and reduce the number of searches users perform, and that this reduction more than compensates for the token cost.

In other words, users get more value from their Kagi subscription, trhough AI grounded in Kagi search, and we're able to offer that extra functionality without increasing costs.

Obviously, this varies by individual, not everyone searches or uses AI the same way. But overall we've found we can add such functionality to all subscriptions without raising prices.

JR

Or, simply offer the option to trade the AI tokens for more traditional searches.

Kaguy

JR Oh, I second that. That could be interesting.

Kaguy

Luis Our hypothesis is that AI features like Quick Answer or access to Kagi Assistant have utility and reduce the number of searches users perform, and that this reduction more than compensates for the token cost.

So this is a hypothesis, you don't actually know if that's the case? This reduction might, for example, not be happening at all, or the results given by LLMs might not be satisfactory enough which leads to a higher usage of tokens?

Luis In other words, users get more value from their Kagi subscription,

While I appreciate the gesture, your concept of "value" may not be the same as several customers who are asking for said value to not be added to their subscriptions.

Luis But overall we've found we can add such functionality to all subscriptions without raising prices.

Again, is that a measurable fact or still a hypothesis?

Luis

Kaguy

So this is a hypothesis, you don't actually know if that's the case? This reduction might, for example, not be happening at all, or the results given by LLMs might not be satisfactory enough which leads to a higher usage of tokens?

Yes of course 🙂 sorry my previous answer was ambiguous. We validated this a long while ago with Quick Answer, and again earlier last year after enabling Kagi Assistant for all subscribers (before it was available to Ultimate subs only)

While I appreciate the gesture, your concept of "value" may not be the same as several customers who are asking for said value to not be added to their subscriptions.

That's fair. We know that AI (specifically generative AI such as LLMs) are a disruptive technology. We've published and transparently discussed our approach with the community. Our goal is to ship value to our customers as quickly as possible. While some users don't use Quick Answer, it's a valuable feature for many others that respects our integration principles (laid out in the aticle linked)

Kaguy

Luis Yes of course 🙂 sorry my previous answer was ambiguous. We validated this a long while ago with Quick Answer, and again earlier last year after enabling Kagi Assistant for all subscribers (before it was available to Ultimate subs only)

Sorry, are you're saying your hypothesis was validated or not?

Luis Our goal is to ship value to our customers as quickly as possible.

This is another can of worms, but I thought your goal was to humanize the internet. A customer can decide for themselves what they think constitutes more or less value to their subscriptions, I feel.

Pleonasm

Different customers have different costs to Kagi, and therefore different levels of profit at the same level of revenue. For example, if Alice and Bob are both on an Ultimate plan, and Alice uses 200 searches per month while Bob uses 400 with all else being equal, then the profitability of Alice will be greater than that of Bob. The same principle applies if Alice does not use any of the AI features of Kagi while Bob does, with all else being equal.

If Alice and other similar customers were permitted to donate a portion of their higher profit to a FOSS project, then the profitability of those accounts would diminish. As a consequence, Kagi would have less money to support the continued development of the company’s products – unless it raised the subscription fee for all customers to offset those donations. Are advocates of the “option to disable AI features & donate money for FOSS instead” willing to endorse those outcomes?

Overall, the larger issue is that it may be suboptimal to micromanage a subscription business at the level of individual customers. The more useful way to view the business problem is to think about managing a portfolio of customers, and to administer subscription costs and benefits across that portfolio to achieve the overall revenue objective of the company. Viewed in this light, there is no “excess profit” from Alice that is available for donation.