I know it may sound like a lot (and definitely is a lot for some folks), but when I think about it as literally being the window to a huge amount of knowledge I take in, the fact that it's way less than Adobe products, and similar to things like GitHub, Figma, Notion, Airtable, etc. feels pretty reasonable.
> ANSWER
> I am an AI language model powered by OpenAI's GPT-3.5-turbo. My primary purpose is to understand and generate human-like text based on the input I receive. I can perform various tasks, such as answering questions, providing explanations, and even writing code snippets.
Looks like it's not on v4 yet.
It's a pretty big problem in internet tech, actually. Ad revenue has poisoned the well, creating a situation where it's really, really hard to create a compelling product that isn't based on advertising, because you end up having to compete with "free" alternatives. Thus from a consumer's perspective the cost of the product needs to be in the added functionality over the alternatives, since they see your base functionality's perceived value as $0.
It's gotten to a point where I don't think it's fixable at a grassroots level.
Brave Search also allows you to customize your search results with "Goggles". These are plain-text files that you can host on Github or Gitlab and use to boost or remove specific sites from your search results. For example, you could create a Goggle that removes "quora.com" from your results while giving priority to StackExchange sites and Reddit. You could also create another Goggle that shows results only from a predefined set of cooking sites.
Finally, multilingual users may find it helpful to create different search shortcuts with the `country=` parameter appended to them. For instance, if you add `&country=fr` to a search URL, you'll get French results.
I suspect there's a big difference between an ordinary user and an ordinary Kagi user.
I guess I don't search nearly as much as I thought I did. Still worth the $10/mo to me.
It's like anti-marketing. Target the paying customers most likely to talk a lot about your product – power users with heavy search workloads – and do stuff that makes them unhappy. And that's you, I guess: I'm reading your comment because you're the sort of person who *would* write HN comments reviewing their new search engine. And now I'm unhappy with it too, by contagion.
(30 searches/day is for illiterate gibbons)
They also made it very clear on that early-days pricing that it was essentially arbitrary as they just had no idea of how the economics of their business would develop, and that they'd need to revisit it. So it really seems unfair to use that as an anchor.
They're also being extremely open about everything. Anyone can join their Discord, and there is a channel there about pricing that had been going for months, prior to the change, where people could contribute to discussions and meaningfully get their grievances heard if they had any.
- Stay at Professional - $10 + $49.5 query cost = $59.5 total (495% price increase)
- Use grandfathered - $10 + $37.5 query cost = $47.5 total (375% price increase)
- Use Ultimate plan - $25 + $0 query cost = $25 total (150% price increase)
With absolutely 0 guarantee ultimate plan will be ultimate. Would I pay for search engine 25$? Of course. Does Kagi delivers 25$ of search quality - no. As I mentioned I did metrics. 50% of the queries had to be moved to other search engines, so Kagi simply didn't deliver.
I didn't find that weird at all: They were overly optimistic, both on how many people they'd be able to recruit for a paid service, and on the economics of how many searches these people do.
I'd be willing to bet that those skew heavily in the direction of power users and knowledge workers using search professionally, so they probably do a lot of searches. I was really surprised when I read my own numbers. In the survey they asked me to fill out, I wrote that I'd expect I'd be doing maybe five searches a day. Turns out the number is actually in the range of 30-70 most days.
One data point that seems to support that is that Kagi previously didn't take any outside investment, with the whole thing being bankrolled by the founder who had had a prior successful exit. But at the same time when they announced the pricing, they also announced they'd be willing to take on board small investments from enthusiast users. So it's looking like they really are starting to get worried about money.
I also indicated in the survey, that, for a really good search engine, the amount of money I'd be willing to pay would be a multiple of what their actual pricing is now. A good search engine is an investment that could pay phenomenal dividends for me, allowing me to make better decisions for my business by having better information as input. It allows me to waste less time, when searching for solutions to technical problems. It allows me to connect to the right people. So, I don't want to think of search as a commodity, like electricity. I think that, as a knowledge worker, I'm really only ever as good as the search engine I'm using, and I just want to hand somebody a pile of money to make me the Ferrari of search engines so I can have a competitive advantage against all those people still riding the bus. (The bus is full of billboards and paid for by people who never want the bus to reach its destination, so they can keep the passengers' eyeballs focused on the advertising).
It sounds a bit like the use case of ChatGPT/BingAI. Depending on your specific use case it might be the wrong fit - especially if you want to look up PI - however it is excellent as a rubberduck/idea provider.
You can also directly ask for the sources btw. In the end LLMs with access to the internet can be an abstraction to search engines. If it is good can only be assessed once you actually tried as it heavily depends on your use case and your approach.
LLMs are just a tool like everything else.
I want completeness (i.e. if source material is being skipped, I want to be the one to make the decision that it can safely be skipped), and I want consistency (i.e. I want to be able to double-check that the same set of sources wouldn't also support drawing conclusions that contradict those drawn by the AI).
In other words: I'd still want to use search as a product, and I'd still want someone to solve the search problem for me, even if question answering were a solved problem.
1. Force you to signup to see the content (pinterest, I'm looking at you).
2. Copy-n-paste SO threads onto a page that is 90% ads.
I expect that ChatGPT written content for sites that are 90% ads will soon make it difficult to determine the useful sites.
HN might possibly be the wealthiest collection of people on the internet. And if we’re complaining about $10-20/month here, there’s certainly no hope for the mass market.
if you know what an SRE III is, and where those exist, chances are you're in the global 1%, and top 25% of US earners.
and yah i get not everyone on HN is STEM, and not all STEM is super-lucrative, but the parent poster is directionally correct.
I'm happy to pay the current $120 I do for Kagi, but that's my upper limit.
I do have some web tools I pay $250 a year for, and they are significantly more complicated than kagi.
With Kagi's new pricing, I will be paying >$300 a year. That is too much. They aren't THAT much better. That is literally more than my beefy dedicated gaming server in a datacenter!
Hell, I probably built my entire career off internet searches.
In theory, HN people should have the most to gain from a better search engine, and be the most likely to pay for one. Yet, to your point, the alternatives are free and good enough.
The problem is, what the mob moralistically claims they want on HN (when upvotes/downvotes are a concern), vs. what those same people actually do in the real world on an individual level, are two separate things.
The lesson is, things evolved the way they have on the internet for a reason. It's not some "evil" companies shoving these ad-supported business models down our throats. We have all voted with our eyeballs and our (lack of) dollars, and will continue to do so. We get exactly the products we deserve.
The early resistance towards a paid model is to be expected. But if we can learn from the example of YouTube Premium (25M+ subscribers for a product available for free), there are millions of people ready to pay for search today, we just haven't reached them.
No reason both can't be true. evil companies can shovel ad-supported business models down our throats while the few paid options which have emerged haven't offered users enough value to justify the fees, being forced to be logged into an account, or having to worry about how many searches you make in a day.
That's a lot compared to my modest 20 searches a day on average. So I'd guess you're an outlier, since I do sit within their "30 searches a day" claim. I use Kagi on desktop while working, and on mobile outside of work. I've never hit my 1k searches/mo limit, but was honestly worried I would (and set hard limits to not exceed my limits). I guess I don't search as much as I thought I did. I really like Kagi and think it's well worth the $10/mo.
(I'd really like to hear how you're hitting that search volume, if you're open to share? I can't imagine hitting a search engine that many times a day, so I'm quite curious.)
An leisure example - I was checking various fidget toys. My ~10 minutes "research" took 15 queries like: "spring toy", "fidget stick", "worry beads", "kendama", "baoding balls" etc. (on topic: infinity cube is quite cool)
Another one would be preparing instrument study sheet with around 20 music instrument where I searched both for images & audio (~50 searches) - e.g. "acoustic guitar sound sample" & "acoustic guitar".
Those are simple, professionally I sometimes go quite wide to find hook point and then dig from there. 20-40 queries spread across various search engines.
I am aware that I'm outlier, yet thinking about usage in this case is a mistake for me.
The whole explanation of the recent “we want to stay afloat” was weird, unexpected (price doubling with my usage, to 20$/mo) and felt untrue (ordinary user doing 30 searches per day?). I felt like it was written by Cut-Me-Own-Throat Dibbler - made sense but also very abstract and shifting responsibility to the customer.
In the end I ran stats how many of Kagi results were on spot (i.e. weren’t repeated in other search engines). It was around half, probably ones that I “didn’t pay for”.
I was ready to move back to DDG but an acquaintance recommended Brave search to me (which has its own index). I was reluctant to use it due controversial opinions about them. Not only I found it more pleasant to use, but also the results are (in my opinion) slightly better than Kagis.