Perplexity launches AI agents for work | Hacker News Reader

Perplexity launches AI agents for work

perplexity.ai

64 pointscybermango54 comments

It's crazy how they were THE AI to go to for researchers and anti-hallucination stuff. And how quickly they became irrelevant.

But Sam Altman said back in the GPT-3.5 days that exactly this will happen to all GPT wrapper startups. [0]

[0]: https://x.com/ai_for_success/status/1779930498623742187

Well, the circle around "wrapper company" is quickly expanding.

https://xNotMe.com/ai_for_success/status/1779930498623742187

> to go to for ... anti-hallucination stuff

In theory maybe, in some tentative hope that the RAG would patch the hallucination problem - but the LLM showed such a low effective equivalent IQ that its usage was always incredibly frustrating. (It could have improved recently - I am not sure. Some tests have indicated it as especially dumb even not long ago.)

Yeah, it seemed like snake oil to me. It never worked reliably.

Oh yes, I don't disagree, but in my recollection, the pitch and the hype was always about how they never create fake sources and fake citations, at the time when that was a problem.

> go to it

You would go to it, but results showed much more work was needed - also in RAG. Which became a general focus in the community, for the many problem it may solve and the many it could help fixing. Some found better formulas for RAG, others swamped. Perplexity made it a flag, but in a factual large company. And the problem with "fake sources and citations" is still there, while mitigated in the full complex systems - hallucination is still a benchmarked effect in LLM ANNs, and blunders from excessive trust are still recorded.

I still remember an HN post back when Perplexity was still the current hotness where the top comment about whatever AI news was analyzing MSFT stock with price history and key milestones, and cited Perplexity as their source. With some meta commentary about how easy research is now that tools like Perplexity are available.

Literally every cited number was flat out wrong. Some people replied to point out the mistakes but as I recall no one called out the Perplexity elephant in the room specifically.

I'll have to find it again, it was such an absurdly ironic scenario it felt like performance art or something.

YOu need to test and use it yourself.

The differences between Opus 4.8, Opus 5.0 and Opus 5.5 are huge.

>they were THE AI

Lol, that's not true.

They always were a "literally who". A lot of paid propaganda on blogs and even shills that were coming to sites like this, on Twitter, etc... were writing stuff for pennies per post (lmao) about how much they used it for this and that but it was all lies.

But people were not using them, regular non-tech folks have ZERO awareness of their brand.

There was a very brief period in the early days where Perplexity actually was pretty useful. I believe I even had a paid subscription to it for maybe half a year at that time. The main value it had then is that ChatGPT didn't yet incorporate web search, or early on after it did its searches were will really limited. There's no reason for Perplexity to exist now. I'm confident they and their investors are keeping it going mainly because no one wants to be the first major AI startup wipeout.

We used to have a lot of lectures of various kind, ranging from short 1-2 hour presentation of 'what's best and what to use' to three-day workshops on academic AI tools, etc. And Perplexity was always mentioned, alongside Scopus AI, Google Scholar AI, etc...

Besides some testing, I've never used it too for different reasons - I don't like my chat history and questions stored at some third party vendor that's just a startup.

For me, I think Perplexity and Siri/Apple are the biggest fumbles so far… They that the users and I have no clue why they could not adapt

Is perplexity still relevant?

Hardly, but I applaud the effort they do to remain in existence.

They might get relegated to niche use cases, or pivot like Midjourney. I'm glad they exist, even for the sake of fighting against the duopoly that's just forming with OpenAI and Anthropic.

Perplexity can fight Gemini for that 3rd spot. Gemini is kind of fumbling it, so they have a real fighting chance, imho.

Why is Gemini fumbling it? I use it for most of my personal projects and it's great for those.

I get better quality output from Perplexity than Gemini.

That said, I stick to Claude in the terminal for 99% of my LLM usage.

Is this a gemini vs Antigravity-cli thing? I use agy as my cli at home and it's as good as Opus/Sonnet for me -- but I haven't tried very taxing tasks. I use Aragon/Flash. I'm not modifying large code bases, more like vibe coding throwaway projects at home.

I too get better flavor from dog turds than rat turds.

Not disputing this.

That said, LLMs can be fun toys regardless of the hype.

Toys, sure.

It's usually not a good idea to build gas pipelines directly to your toys in the middle of a climate crisis.

Again, not disputing this either.

I use Gemini it's bundled with my phone contract. Its privacy is nightmare there is no opt out.

Yeah, but all of the LLMs are a privacy nightmare, as is any Google product.

From and investor and "market" perspective, the only AI that matters is the one built into your phone, TV, car, and all your smarthome appliances. Google and Apple completely own this space.

If anything Anthropic and Open AI are at a severe disadvantage. Open AI at least provides backend model services for Siri. No idea what Anthropic does.

> From and investor and "market" perspective, he only AI that matters is the one built into your phone, TV, car, and all your smarthome appliances.

They can still be disqualified culturally speaking, as in, yeah every time you unbox a new smartphone it forces you to accept terms and conditions, yet you never open Gemini or Siri. That's great for the someone that will get that promotion and pay raise.

An investor sees that app installs are increasing and that's "great for business," yet it's shallow because the real app usage isn't that high.

"How come we have so many new installs but Claude is far more used?" They have all that telemetry, they should know that for certain.

Google and Apple can force installation and even terms and conditions agreement, but they cannot fake user commitment.

> Google and Apple completely own this space.

Microsoft bundled Internet Explorer, and once it became technically inferior (IE7+) to others, Chrome and Firefox ate the market share of it. Then it was Chrome all the way in.

Internet Explorer became "the browser you loved to hate," and they even did a campaign on that motto.

The technical debt was so heavy, they had to start over with Edge, and it still doesn't gain market share as it is expected. Just because it comes deeply ingrained in the mobile OS that still doesn't grant them ability to win the hearts of the users.

I'm one of those that delete the Gemini app every now and then (it becomes auto-installed with each OS update, it's hostile behavior).

Gemini and Siri still have to be good even if it's forced on our devices. We already know Siri lost it[0]. Siri is comically bad.

[0]: https://news.ycombinator.com/item?id=50104465

> Open AI at least provides backend model services for Siri.

Do they? I thought that was Google/Gemini now.

https://www.cnbc.com/2026/01/12/apple-google-ai-siri-gemini....

I sill have a ChatGPT extension option in Siri settings on 27.2; perhaps the default Apple Foundation model is Gemini-backed and the ChatGPT option still remains

Funny Friday

Everything about that sounds like Google got it figured out for themselves.

The CEO just needs to do more CNBC appearances and they're good.

I do not understand how you can conclude that they both are even competitors - along any dimension

Perplexity has now become a speciality tool used for a few use cases primarily via third party models (unsure if they are still continuing to develop their own mdoels and where they rank)

Google on the other hand is still actively trying to develop SOTA models (or at least 3rd best) and also is widely used amongst cost sensitive projects.

I see no comparison between them worth putting them alongside each other.

They have users and revenue in the hundreds of millions so yes. No affiliation. I don't use their service either.

I also think that this is going to be very useful and many startups are entering this space of agent teammates with persistent memory.

Absolutely! I use it all day as a fact finder and search engine. It's a much cleaner interface than Google, no ads or spam results, and it does basic analysis quite well. Recently I've been asking it to "Tell me the downsides of living at the following location..." and it does a solid job of synthesizing common complaints and it has memory, so it can target that to what I've said before.

Nope. There is no reason to use perplexity.

I only stopped using Perplexity when the kill-sticky (login-workaround) stopped working correctly.

Early 2023 - August 2026, I used their rudimentary LLM before making any major purchase – seems like they really dropped the ball by both locking out users (log-in requirements) and not exploiting their early-Top10-LLM status to leverage into corporate ad influencing (again, which is how I primarily informed my own purchases, until login requirement).

I run most of my LLM locally at this point, but it would be great to have perplexity be the autistic shopping tool it once was. Ce la vi

I personally enjoy the news feed part of it. It cleanly summarizes the major global news stories across several sources into 3 bullet points, and it doesn't show you any ads disguising as stories (eg product reviews, tips, etc)

Perplexity seems desperate. I used to be a subscriber, but they've been making their service worse to save money using tricks like "forgetting" that you picked a smarter model and switching you "accidentally" to their cheaper (and dumber) one.

Claude and ChatGPT basically offer the same functionality now, but with better models and more features.

When I started using Perplexity when they started out, it was such an awesome product. And they held that position for some time.

For some reason, I feel like it has gotten worse than it was. I don't know if it's just me.

Maybe chatgpt/claude just got increasingly better. Yet still feel they should have been able to keep up with it at least? but now they are just so much behind in terms of user experience and general product quality.

They substituted models quite often, and imposed low usage limits for users who had been onboarded for free. Supposedly it was for specific adverse situations such as heavy usage or overall load. [1]

[1]: https://www.reddit.com/r/perplexity_ai/comments/1orar1a/upda...

I remember asking perplexity about the instructions for upgrading an open source app from several versions back to the latest release. I had hoped that it would search the relevant github, stack, overflow, and blogs. Instead, it hallucinated some instructions from its training corpus, which was incomplete and outdated. When I pushed back, “no, I want you to search for and read the latest instructions” it told me it couldn’t search for newer data and what it had already given me was as recent as it could provide.

If their tool can’t follow open source README files, why would I trust it with other research?

Does anyone know if they have done secondary sales / tender offers for employees?

They’re going to be sold for parts at some point at a valuation where common shareholders get zero.

I remember a lot of engineers joining with the promise of easy money and stock options “worth” millions. Now the company is basically irrelevant.

Who?

Didn't they make a computer thing like a year ago? I think it was called computer. (Right around the time OpenClaw was big?)

Edit: https://www.perplexity.ai/products/computer

I forgot Perplexity even still existed. They went from all the rage to nobody talks about them anymore.

Honest question… what IP does Perplexity have that gives them any moat whatsoever left in this increasingly commoditized market?

Name ID. They were around early and got their name out there. If you don't keep up with the blow-by-blow of the tech and the surrounding market, you may not know that there are better alternatives available.

>I forgot Perplexity even still existed. They went from all the rage to nobody talks about them anymore.

A lot of the early hype was driven by how liberally they were handing out free Pro memberships. For a while, every telco, bank, etc. was handing it out as a free promotional offer. I don't think anyone I know that raved about it ever gave them a single dime.

Their product was definitely cool for a few months, but they had zero moat to begin with (considering how quickly the frontier labs caught up), and the conversion rate on those free sign-ups must have been abysmal, especially after they nerfed the free Pro subscriptions on both quality and quantity dimensions, and their own "politically neutral" DeepSeek R1 distillation didn't pan out.

What moat does anyone have? Google wrote the "we have no moat" memo and they're literally Google (better hardware, better data, billions of users...)

ChatGPT has the name, and Anthropic has ... the writing style is less cringe? (Well, it used to be, anyway...)

If there's no moat then there's no need to be skeptical about a company with hundreds of millions in revenue. Their brand resonates well enough for that.

Since someone coined Microslop, may I give you “Poorplexity”

:s/r/p. iykyk.

rip perplexity

what is it that these lot do again?

Reply on news.ycombinator.com