Provide a plan without AI functionality

Thibaultmol

luked I don't think I've seen improvements to Teclis once in changelog?

Tbf, I don't know if 'Teclis' itself is being updated BUT Kagi's own indexing/crawling efforts are being actively worked on with various people behind the scenes.. it's just .. behind the scenes work to setup the infrastructure and such but none of those things show up on changelogs.

Various people on working on Search, maps (both non ai related things), some things just take time.
Something that was launched recently to Search: https://blog.kagi.com/slopstop

user_7575

What does your feature entail?

I would like if Kagi Assist would be optional when I pay for Kagi. Maybe Kagi Search is available to pay, but Assist is an optional add-on or bundled in a higher tier instead of being bundled with all plans. This is for users like myself who don't like AI¹ and, not only would like to not use Assist, but also don't want to pay extra money for Assist when I have no desire in using it.

Right now, $10 will get you unlimited searches and Assist. Instead of this, maybe the plan could be like $8 for unlimited searches, and if users wanted to, add an additional $2 for Assist.

¹ I know there's genuine use cases for AI and I appreciate Kagi's AI philosophy, but I'm so sick of the technology being awkwardly shoved everywhere. It's why I made this suggestion in the first place.

What are the exact ways that you see a user using your proposed feature?

Again, for users who'd like to distance themselves from AI anything, they get to do that.

They also get to save some money, too. It's not just the initial plans users have to pay for, users may have to also pay taxes and foreign transaction fees, too

xenofem

Thibaultmol Tbc Kagi isn't trying to force ai onto people

When I use an ad-funded search engine and it shows me a sponsored link, it technically isn't forcing anything onto me either. If I'm not interested in what's being advertised, I don't have to click the link. It's still an intrusion, though—it's taking over my screen space and my attention to market something at me that I don't want. I feel the exact same way about "Quick Answer" buttons and all the other AI functionality in Kagi's UI. It's not just that I don't want to use it—I don't want it offered to me at all. If I pay money for a search engine, I don't want it showing me unsolicited advertisements for things that I'm ideologically and professionally opposed to, and I don't want to have to set up my own ad-blocking tools (custom CSS) and keep them updated whenever Kagi makes any changes to their layout. That's my position here.

TomFryers

I had been considering subscribing to Kagi, but am unlikely to do so unless there’s an option that excludes LLM features. Paying for a search + LLM plan counts as paying for an LLM in my book, and I’m not comfortable doing that.

It’s worth pondering the idea that it would cost more not to have LLMs. It implies that the LLM features have a negative price, which is curious. In fact, there’s a lovely English word for something that you have to pay to have somebody take away: rubbish.

maddyunderstars

I would like to express support for Kagi to focus less on LLM technology. Ideally, I would like to see no involvement with OpenAI or other LLM providers.

I understand my subscription cost has not increased due to it, however it has been used to fund it's development and integration with Kagi.

I would like to see Kagi focus more on (I might move these to their own thread, eventually):

Search widgets e.g. better unit detection in calculator
Kagi Maps e.g. directions, and the ability to use it in the Kagi mobile app
Additional regional integrations, e.g. public transport for Australia in Kagi Maps and search

Some in this thread have mentioned that work in these areas does occur, but that it's simply not worth mentioning in the change log. I disagree and would love to see a technical change log or occasionally blog post about inner workings.

Edit: I don't have issue with general AI features. If it is ethically trained, has little environmental impact, and little resource cost (e.g. not enough to cause a RAM shortage) then I'm fine. Ente's on device ML for face detection is a good example, as well as Kagi's AI image filters. LLM companies like OpenAI do cause issues here, however.

Kaguy

As a new customer, I would like to express that I also wish for Kagi to stop dabbling with AI so much. I'm paying for quality search, and no matter how much one may say the costs associated with the AI functionalities are non-existent or based on a user's individual usage of said features, I simply do not believe no resources at all that could go into improving the search aren't being funneled into AI features. One would assume you'd still have to pay something for model providers, neoclouds or pay developers to make your own, not to mention development time to implement functionality.

I find features such as "Kagi News" particularly egregious, as that's just LLM summarized news which are often wrong as one can gleam from just visiting this very forum, yet it poses as a news aggregator. Tweaking an LLM on endless user feedback so it won't relay faulty information about real world events feels like an unnecessary step in order to support a fundamentally flawed feature, such as the LLM search results if you type a "?" at the end of the query. I've even seen discussions here about Kagi's models supporting image generation, which is not only something I'm particularly staunchly against, I fail to see how that's in any way related to a search engine, let alone your mission statement of humanizing the web.

I get the Kagi team seems to be really excited about AI. If I'm not mistaken, the company started as an AI startup. But this feels like another case of shoving a feature into a product that really has no business being there, and I have to wonder how all those LLM-reliant features will go on working once the speculative bubble we find ourselves in bursts and the real costs involved with leasing and upkeeping LLM models creep up.

I'm uncomfortable with those features existing. As evidenced here, other customers are as well. I appreciate I'm not forced to use them, but their very existence ties Kagi as a whole to a system many are not keen in participating, and if you wish to humanize the web and champion the good old "you pay us, we give you a good product", perhaps you should take your customers' comfort into consideration and do the "extra work on billing" instead of continuously reasurring "no one has to use it, you can just hide it, some people like it", "there's no cost to you", etc. I do not want it. I will not use it. I do not want to suspect my money is going toward even development time and resources to implement such features when other, more important things could be worked on. That was not the pitch.

Kaguy

Vlad I'm sorry, I know this is an old message, but I feel like I need to respond to this specifically because the whole argument feels particularly egregious.

Even then, I believe this is the best way to consume news. Have you tried reading a news website recently? Between ads, tracking, low-quality journalism and opinionated reporting, it's impossible to consume. The way we approach it in Kagi News not only removes all the noise and sticks to the facts, but also offers perspectives on events from diagonally different sources. That's all I care about from journalism, and it would take hours to manually synthesize for just one event.

An LLM summarizing news sources does not remove any noise, if anything it adds to it. The act of summarizing anything correctly involves human input in understanding context and thus what is important and core to the information being presented and what are details that can be left out. To honestly believe an LLM is capable of doing that consistently by continuously guessing what the next most likely word is is observably, demonstrably untrue, especially given the nuance of human politics that generate the news in the first place. Daily news even.
How does an LLM remove the "noise" of low-quality journalism and opinionated reporting when both those things are necessarily present in the sources it is pulling from to summarize? Is Kagi's model somehow omniscient to the point it can stick to facts it did not observe nor experience by summarizing what other humans are saying by their already subjective and opinionated lens? That sounds extremely unlikely. Offering links to all perspectives of any given discussion is what a news aggregator does, and one can read these different perspectives for themselves, this technology of "looking for different news sources" has been around for a while and I fail to see how an LLM adding the extra step of summarizing is any real benefit, especially when the summarizing is often faulty.

Vlad Point to an actual example of bullshit created by Kagi News.

Just open the feedback page for Kagi News right now. Several complaints about untrustworthy sources, innacurate summaries, context pictures being show below headlines in ways that lead to misunderstandings of interpretation. Users having to consistently point those very complex and subjective things out for you to somehow patch out does not corroborate your claim that the process of following the news can be reliably automated. It does not have intrusive ads, I'll give you that.

Turns out journalism, the process of relaying information and the inherent human bias present when reporting the facts as observed by our collective subjective experiences in an interconnected world with all sorts of sets of incentives are all actually extremelly complex and nuanced things that can't be reliably summarized by an algorithm that has no capacity to truly understand the context and meaning of what it is regurgitating, nor can it present anything in an "objective" way considering the pool from which it pulls its "understanding" is already biased to begin with.

You can't expel bias, subjectivity or any otherwise perceived human faults from the news (or anything as grand in scope, really) with an LLM, Vlad. You are welcome to try, but I'd be extremely glad if you did so away from the product I'm paying for, preferrebly with a different mission statement from "humanizing the web", because there seems to be a fundamental misunderstanding on what the word "humanize" means.

This bullheaded insistence on implementing LLM features (note I'm specifically referring to LLMs, before the "AI is other things that have been used forever" argument turns around, I know, everyone knows) to what's supposed to be a search engine makes me increasingly distrustful of the choices the company is making, not to mention the claims regarding how much it's costing you or where exactly my money is being sent. You and your team seem to be really into LLMs, it is pretty fascinating tech, but I'm not paying you pursue your personal interests, I'm paying you for providing me with good internet search, and the last thing I need is a large language model trying to do my search for me or resources I contribute to being directed toward something like Kagi News, which I unironically believe to be extremely dangerous, no matter your intentions.

Pleonasm

While I do respect the sentiment of those who prefer (1) not to use and/or (2) not to fund AI through their Kagi subscription, having a Kagi subscription option that omits access to AI features appears to be an emotional rather than rational reaction, from my perspective (and I recognize that others will disagree).

Kagi does not force a customer to use the AI capabilities in its product, and ignoring those capabilities is not burdensome.
Suggesting that Kagi subscriptions pay for (and therefore support) AI activities seems to be a reflection of the “mental accounting” fallacy discussed in the behavioral economics literature. More specifically, the money received by Kagi from subscriptions is fungible. The received subscription revenue is simply a single account of “neutral” money. Kagi’s profit may change per user based on incurred expenses to support that person’s usage of AI, but revenue and profit should not be confused.

Gredharm

Pleonasm Correct. I've noticed this feedback forum has a disturbing amount of emotional reactions compared to valid feedback. It's probably only 20% or so, but it really stands out.

There's a lot of good feedback, like bugs, UI suggestions, feature suggestions.

But the ones that get the most traction seem to be the emotional, personal matters that really have nothing to do with anything technical or relevant. Like complaining about what search providers are used, or AI, or other references sources.

maddyunderstars

Pleonasm Our subscriptions cover the cost of Kagi's development, including their team's salaries, office space, equipment, and any other costs the company has. The development of LLM tools in Kagi directly comes from our support via subscriptions.

I agree that the use of the tools is optional and does not directly increase our cost. However, the developer time used to create the functionality is supported by us.

I would like to see Kagi developer time and other R&D used to improve non-LLM tools.

Kaguy

Gredharm Whether customers are being emotional or not hardly matters. I'd even argue what you are calling "emotional" is anything but given the current world circumstances, but I just wrote a novel of a comment nobody will ever read and all arguments for both sides have been exhausted.

The matter of fact is that a sizeable portion of customers are averse to AI and would like the option to have no AI functionality, no matter what you personally think about the rationality of this aversion. It should be possible to do so either by way of another plan or providing a button to disable any and all access to it. This is a feature consistently asked, and all we get is "just don't use it" or "it makes no sense to allow you to block yourself from using it" or "it adds value to your subscription". How about "just do the feature we're asking for"?

Dustin

Kaguy The matter of fact is that a sizeable portion of customers are averse to AI and would like the option to have no AI functionality

I don't know if this statement can be substantiated by any evidence. 63,000+ paying customers now and I don't know how much a couple dozen in this thread really represent the views of the entire customer base. Only Kagi knows, I suppose.

The reality is that most people will not give feedback for or against. Most will not say anything at all. As is often the case, one can absolutely find a very outspoken minority who by dint of social contagion or moral outrage will take the time to give their feedback as you have done @Kaguy. They are customers, too, and their feedback I'm sure is well considered by the Kagi team.

Kaguy

Dustin By the same token, we also don't know if Kagi's continued statements that adding all the LLM features has attracted customers who otherwise wouldn't subscribe. We also don't know how many of these 63,000 even like AI that much in order to justify such focus.
This very forum exists for customers to give feedback. If by your very premise very few bother to do so, then everyone here is a loud minority whose feedback should be considered "relevant" only with a grain of salt, which obviously defeats the purpose. Under this framework, literally all opinions here can be argued to be "loud" minorities by the virtue of being noticeable and not being 23000 people saying the same thing.
Also, while I'm sure that was not your intention, I'd appreciate avoiding saying people arguing for this position do so out of "social contagion" or "moral outrage" as primary factors. Like I said, there are logical reasons for having these opinions, and claiming this obviously diminishes the validity of the complaint.

Pleonasm

Imagine that the Kagi benefit plan provides a very generous reimbursement of medical expenses to support employees who wish to have an abortion. Given that a sizeable subset of Kagi’s customers may be sincerely averse to this practice on ethical grounds, should Kagi offer two versions of its subscription: one that ‘includes’ the cost of this benefit, and another with pro rata amounts subtracted from the subscription fee to accommodate customers who have this aversion?

I'd say that's obviously not very analogous to the situation since abortion is something much more dependant on literal government policies than users of a service voting with their wallets, but for argument's sake (meaning pretending abortion follows the same set of incentives as Kagi), "yes, I think they should do that" is my answer to your question. Unless we agree something like abortion is a deeply different thing than a premium search engine, of course. But that would destroy the analogy.

Additionally, should there be a button on the Kagi interface that automatically disables access to any and all search results that have a favorable view toward the practice of abortion?

There already is in the form of lenses and user's personal downranking of search results, for example. If a user wishes to deliberatly personalize their search into a bubble, they already can and should be able to do so. Not sure where the moral quandry is here.

Clearly, it would be easy to create a long list of subscription options that include or exclude costs and features associated with some aspect of Kagi as a company or as a service. Managing a business in this way, however, would be less than optimal

Only if there were several groups of users demanding extremely specific carvings of every aspect of Kagi, which to my knowledge is not the case. Even if we assume it's a "loud minority", what's being asked is quite clear and sweeping: allow me to block AI functionality from my experience. It's a simple request with an actionable solution. The problem you speak of is not happening right now nor are there signs it ever will.

Pleonasm

On the subject of “paying” for AI through the Kagi subscription, consider this hypothetical example. Imagine that the Kagi benefit plan provides a very generous reimbursement of medical expenses to support employees who wish to have an abortion. Given that a sizeable subset of Kagi’s customers may be sincerely averse to this practice on ethical grounds, should Kagi offer two versions of its subscription: one that ‘includes’ the cost of this benefit, and another with pro rata amounts subtracted from the subscription fee to accommodate customers who have this aversion?

Additionally, should there be a button on the Kagi interface that automatically disables access to any and all search results that have a favorable view toward the practice of abortion?

Clearly, it would be easy to create a long list of subscription options that include or exclude costs and features associated with some aspect of Kagi as a company or as a service. Managing a business in this way, however, would be less than optimal

carl

Kaguy If by your very premise very few bother to do so, then everyone here is a loud minority whose feedback should be considered "relevant" only with a grain of salt, which obviously defeats the purpose. Under this framework, literally all opinions here can be argued to be "loud" minorities by the virtue of being noticeable and not being 23000 people saying the same thing.

Well, isn't that obvious? Of course each of us posting here are just voicing our individual opinions, and the feedback forum as a whole is a small subset of users. Most customers don't voice any opinion at all, as already mentioned. So yes, everything here should be taken only with a grain of salt, which I am sure it is.

oImcgMeS1-p0

Personally the thing about Kagi that stands out to me is the privacy and anonymity features, the independent search engine development, the ability to change the prioritization (for instance filtering out AI results and using the fediverse lens I do enjoy) and last but not least the non-ad-supported business model as it rids a natural incentive to prioritize content that aligns with propagandistic narratives dictated by advertisers as not doing so would jeopardize income.

The AI related features are not really part of that list. I honestly would be willing to pay a for a little more expensive non-AI plan. I also found the paragraph where @OnanForumed started with "Of course! It's not the internal indexes..." to be agreeable.

Pleonasm

Readers of this thread may find it interesting that Mozilla recently made the following comment about the company's plans to incorporate AI into its browser: "Firefox will have an option to completely disable all AI features. We’ve been calling it the AI kill switch internally. I’m sure it’ll ship with a less murderous name, but that’s how seriously and absolutely we’re taking this.”

Yet, what practical benefit would accrue to Kagi customers by correspondingly having an "AI kill switch"? Turning off Kagi AI features might provide psychological comfort by creating the illusion that the company and product do not support AI, but otherwise what substantive and concrete benefit would result?

Kaguy

Pleonasm At this point, why question so much what benefit the people asking for this feature will get from it? Even if we assume it to be just psychological comfort, that is by itself a valid motive to ask for a feature in a product you pay for. Why insist on denying your customers this simple comfort just because you personally don't care for it?
If someone really, really doesn't want to risk having their money siphoned toward AI development, they simply won't use Kagi. If they like the company's main product enough to want to use it anyway, all they're asking for is an option to block AI functionality so they don't accidentally trigger an LLM to do something (as evidenced in this very thread) or just because they think the visible options are bothersome.

Why insist on not giving this subset of users what they're asking for? I certainly didn't ask for an assistant and I haven't seen some huge demand from their customer base for one either, but here we are. Giving your customers the option to turn it off completely instead of just ignoring it doesn't seem like that big of an ask.

« Previous Page Next Page »