Hacker News Reader

Is Perplexity, as a business, sustainable? I'm skeptical of these "AI search is better than web search" claims. Not because it's not true, but because of how wildly subsidized these AI things are. Web search doesn't suck because it's an impossible problem, web search sucks because it has to be profitable, and the profitability comes from making it suck. That's not true today for AI tools, but it will be in a few years. What will stop AI tools from being sucked into the same ad-infested black hole once the free money runs out?

For this kind of thing I'm an LLM skeptic, but

People said the same thing you're saying about Google in the late 90s, early 2000s before their IPO (and immediately afterwards). There was a sense in which people didn't really think search -- which seemed more like a public utility -- could ever be profitable. Yahoo and AltaVista and Excite blanketed theirs in ads and junk. The search itself was seen merely as a draw into a "portal."

It's not inevitable that the same thing happen with services like Perplexity. But I do think things are going to get shaken up.

It seems like Google agrees at some level, because they seem to have just given up.

The big difference with Google monetizing in the 90’s is that they owned their platform, which gave them control to build out a search and ad network that scaled.

Perplexity uses frontier models under the hood, it doesn’t own all core aspects of its platform. They’re extremely dependent on frontier models and therefore reliant on underlying model pricing remaining sustainable.

Yeah I don't think Perplexity in particular is all that interesting.

But these days I do do a lot of 'research phase' stuff right in Claude Code. If looking at a technical issue, etc. I'll give it URLs and terms to search for and let it do some research for me. Mixed results.

> web search sucks because it has to be profitable, and the profitability comes from making it suck. That's not true today for AI tools, but it will be in a few years.

Many people pay to use _AI tools_, that already brings in revenue. I had chatgpt plus since very early days, which was 20$/month, I don't have it at the moment because my company provides pro plan to me (and every other engineer) which is probably around 200$/month/user.

Of course, serving a single inference on LLM's probably costs a lot more than a serving a single search on google, but they've already got a solid business model and they won't need intrusive adds _in a few years_ (if at all)

I'm skeptical that a significant number of people will be willing to pay cash money for these products when web search still exists for free, but we'll see!

Perplexity's CEO openly admits that their Comet browser offering is intended to track users for ads. So it seems they will head down the encrapify path soon enough.

Pro tip: get a free account and don't go pro. Their free account is bananas good atm "while supplies last"

I find it's still not that difficult to have that "special power", but you have to adapt your tools. Before the only tool you needed was Google, now you need to know which one to use for each type of request.

I mostly juggle with Google, Kagi, and various LLM (ChatGPT, Claude, Gemini, Perplexity... but the differences matter less).

Try DuckDuck Go, it performs significantly better than Google for me

My experience with DuckDuckGo has been very mixed. I have to frequently resort to Google, especially for error messages where DDG returns zero results yet Google has indexed and finds a GitHub issue with that exact same text in it. The image searches are also really bad. I am sticking with using DDG but I have a shortcut ready in case I need Google and overall I'd really like something else considering it's a thin wrapper around Bing. I'll try out Kagi and see if it's worth paying

It only works well for mainstream search subjects. If something is slightly off the spotlight, it will often fail to find it even with exact name of the page/article in quotes in my experience.

do you have an example of such a query?

> I genuinely think Google is finished as a search engine for the web.

Google Search is garbage, but highly unlikely to be "finished". Millions of people still find it useful, and Google is adopting "AI" on the results page just like any other "AI" web search service. The reason the UX is not good is, first of all, subjective, and second of all, because Google is in the advertising business, and they've found it more profitable to corrupt their results page and deal with any negative feedback, than to deliver clean results like they did decades ago without the profit.

This is a carefully planned, tested, and executed design decision, just like anything they do on the SERP, and not some arbitrary sign that they don't know what they're doing anymore.

The possibility of a new player disrupting the dominance of a trillion-dollar corporation that has built a highly optimized index of the entire web over decades, by leveraging technology that requires vast resources to run, is highly unlikely. Not impossible, but highly unlikely. Google could improve the search UX tomorrow if they wanted to.

> The only problem is that we don't have a solid contender yet.

Sure we do. Kagi offers a much better UX, and I haven't had the need to rely on external results for nearly a year now. I haven't tried Perplexity, but I imagine it could be good as well, depending on the quality of its index.

But these are relatively niche services catering to an audience that cares about these things. The sad reality is that most people simply don't, and will use whatever search engine is set as default in their browser. Which is why being the default is worth paying millions, and is literally keeping companies like Mozilla alive.

This sounds correct to me. Cory Doctorow wrote explicitly about how Kagi uses Google's own index, but just presents the results in a more useful way, which validates everything that you're saying.

> In other words: Kagi is a heavily customized, anonymized front-end to Google.

> The implications of this are stunning. It means that Google's enshittified search-results are a choice. Those ad-strewn, sub-Altavista, spam-drowned search pages are a feature, not a bug. Google prefers those results to Kagi, because Google makes more money out of shit than they would out of delivering a good product

https://pluralistic.net/2024/04/04/teach-me-how-to-shruggie/...

I think youre giving Google too much credit.

They did not expect the public to react to chatGPT the way they did and for OAI to capture mindset and marketshare of Internet Search Queries.

Google deliberately made search worse so users do multiple searches for each query so they can profit off showing more ads:

https://journalrecord.com/2025/02/20/is-google-making-search...

It's not just Perplexity, it's any of the "deep research" offerings on the market really. And the reason why they work so well isn't because of any kind of secret sauce, it's because they have infinite patience to wade through all the bullshit and can thus brute force this without necessarily knowing the shortcuts. It's basically more of the same: https://www.scottsmitelli.com/articles/altoids-by-the-fistfu...

Right? Google is dog caca now. Myself and everyone I know keep getting sent to AI-written garbage nonsense slop websites, or for some reason, to the Hindustan Times

oddly this "caca" felt more visceral to me that most "poop"'s or "shit"'s I've seen in a bit. summoned an image instantly. probably just surprise - good choice!

reminds me of a blackhat presentation of a web crawler

two young gentlemen introduced it as "caca", seemingly an acronym for sth, but they just couldn't help themselves and kept chuckling for next five minutes.

"Caca" means shit in a bunch of languages (at least as a term used with children, but not only, in Romanian, French, etc), that's probably the reason.

"Caca" is more kiki and "poop" is more bouba.

Yeah for some reasons it ranks among the first newspaper any times I am looking for some US news. It feels like someone tweaked the algorithm for money.

I don't disagree - yet their share price hit ATH this month.

That's how you know that our economy is not really working to the benefit of most of us - because all the numerous ways to profit by fucking people up, from dark patterns to mass layoffs, are rewarded financially.

Jaja, dijo caca!

ontopic: This debacle started way earlier than when google decided that the "don't be evil" motto was to be removed, methinks.

I've completely stopped using Google since the ChatGPT search Chrome extension came out. That was, what, almost 2 years ago? More? It simply redirects URL-bar searches to ChatGPT (with the web search toggle active) instead of Google, nothing fancy.

I didn't explicitly decide to stop using Google, it just happened, I didn't need it anymore, just like I went from using StackOverflow daily to never opening it again. ChatGPT with search is just better (at least ChatGPT Plus). Granted, it is noticeably slower to get a first result, but end-to-end it's a much faster way to find your answer.

I have found both perplexity and Claude.ai good enough. Since I pay for claude because of development, why not use it as search engine as well? So maybe the future is multi provider?

> I didn't thought of that as a special power. I thought it was natural, like driving a car, or speaking English.

You’re clearly taking for granted any learnt skills which you have and projecting them to others. A substantial portion of the world population can’t speak English, and I suspect the grand majority of humans don’t know how to drive a car either.

I know you have to turn the wheel to turn the car and that I should keep to the right, but that doesn’t count as actually knowing how to properly operate a car.

OP clearly (to borrow your choice of word) wants to express that they thought people simply learn how to find information by simply using a search engine over time, just like you can become proficient in other activities by repeatedly doing them; and they gave some examples that many people here can relate to as such. I don't think OP wanted to offend anyone who doesn't know how to drive a car, or suggest that everybody should be able to.

So, yeah, things are not always that clear. That's why it's so useful to give people the benefit of a charitable interpretation of their words by default.

The examples OP chose are actually excellent: both of those activities require deliberate practice and years of training (education) and people who don't do either of those suck at them.

Think of the average person you know that had terrible results in middle school and high school or of the average driver, that thinks he's better than 80% of drivers out there.

I'm going to assume good faith on your part. OP is using analogy as a rhetorical device, using things in their own life that would be recognizable and transferable to others. When he says its like "speaking English", it is because he speaks English, but it can be understood that any non-English natives would insert their own mother tongue. 88% of European households and 92% of US households own at least one car. Driving, as something that once learned you take for granted since you do it pretty much daily, is something most people can related to. Even those that cant, are probably intimately familiar with those that do. It isn't something exotic and specialized, like airline pilot or submarine crewmember.

If I say "its like finding out someone ate the last of your favorite candy", it is meant to analogize a feeling of disappointment. If you personally don't like candy, I would expect that you have the ability to understand the meaning and intent and not focus on how the analogy doesn't perfectly align with your individual personal experiences and preferences. If you want to complain about how some cultures don't really eat candy, while others focus on collective sharing over individual ownership, I suppose you can.