Lindy 3.0 launches with agent autopilot feature that operates its own computer and manages logins

Legend. Welcome to the stream. Uh, we kept him waiting. We'll try and move through these next guests quickly. Been very excited to catch up with him. How are you doing? Boom. Welcome. John will be back in a minute. How are you doing, Flo? Good, good, good. How's everyone doing? Great. Great.

Uh, you know, it's uh it's been a slow summer. Not much news in AI. Everyone's just uh clearly taking a breather. Uh, of course not. Um, I know you guys have been busy. What's going on in your world? Yeah, just announced 3. 0. It's been an insane last couple of weeks. Uh, been working on this update all year.

Uh, yeah, I mean, we're just making it way way way easier to create way more powerful AI agents. Incredible. Uh, when did you decide to adopt uh the like versioning for Lindy?

because I it feels smart to like kind of anchor like the team and your customers around these like major updates to the to the product versus historically people might have a product launch but they're not necessarily like signaling like hey this is a massive advancement we're call it's an entirely new version yeah no we've done that from the beginning like 2.

0 you know was like in November last year actually you know even smarter and that's what I want to do next time would have been to name them I want to like Lindy you know like how Apple does liked and stuff I want to give him cool names moving forward yeah I'm trying to think uh we'll come up we'll come up with the right series of of names in the right category do something uh I'm thinking like sci-fi books like what would you think of like a fire up in the dip you know or like the metamorphosis of Lindy Lindy snow crash or Yeah, exactly.

Exactly. Um, that's great. What uh how how are you guys using the new the new Lindy's uh internally? Oh, we're using it all day. Like I think one of the most insane things that it's automated for us is like it's replaced the QA engineer. Wow. It just like so it's getting very meta. Like we all using Lindy to QA Lindy.

And so every hour we have like a Lindy agent that wakes up that tries the entire like core flow like literally signs up with a new account, tries to pay, create a new uh agent, tests the new agents, you have an agent, creating an agent and testing it.

Uh and then if anything goes wrong along the way, uh pings the on call engineer. Uh so we've been using that for weeks. It works beautifully. It's amazing. How do you think about reducing the time uh I guess like time to value for somebody that's coming in to use Lindy? I imagine that's key.

Uh people are so used to like signing up for a SAS product knowing exactly what they're getting and having there be a lot more kind of controls around what the product does. So it's not a sort of like build your own adventure.

Whereas I feel like with Lindy historically you you do like the the customer needs to really embrace like hey I have to figure out how to when you hire a new employee very few roles can somebody just come in and without any ramp up time just immediately start crushing it.

Um and and so I'm curious um what that looks like for you guys. Yeah. Can I show my screen? Does that work? It does. Everything's live. So whatever you share will be okay. whatever into the immutable record of the of the internet. Oh, that looks great. Well, yeah. So, I mean, this is like an excellent question.

Like, that's the problem we've always had from day one is like it's almost too powerful. You have to do everything yourself. And that's why the vision from day one has been like it's just like an employee. Like, it's very easy to set up. You can just talk to it and tell it what you want to do.

And so, now the new app is just a text field. It's like this thing and you tell it what you want to do. It's like, hey, I want an agent that will monitor my support inbox. Then you can like iterate with it like very interactively.

And we have a bunch of templates here for like sales, customer support, personal productivity, engineering right here like the QA assistant is one of them or you know like like a software engineer agent like human resources like a recruiting agent or like a company knowledge base.

So we give you a bunch of templates and then you can just create your agent in like a minute and a half. Are you running into any problems uh just linking up private data sources? is I mean I'm just thinking from like a consumer's perspective.

You have a bunch of uh you know news subscriptions you want to to log in to those news websites and kind of collate a digest for you. That's something it would need a password for. It would get run it would run into uh a number of uh pay walls.

Uh are you seeing a lot of luck with like jumping through login flows or are you getting hit with captures constantly that are intractable? Oh no. I mean like so that is a big part of what we're announcing today is like autopilot.

So basically your agents now have their own computer and if you need to give it a login you can literally take control of the computer and login for them. So let me just pause the screen share because I don't want to show anything crazy.

But does is that is that something that you do like once or because I like when I think of the benefit here I think of like a really longunning like cron job based uh um uh agent essentially something that feels more like a screen sharing still screen sharing. Yeah, your list of passwords is up now. Just kidding.

The API keys are leaked. Uh and it's Oh wow, you did a billion in revenue yesterday. Um, no, I mean, yeah, you love it only once. So, it's like this is I I am screen sharing on purpose. So, this is an agent that said this is actually a little bit of a dirty secret. Like, I don't like LinkedIn. I'm more of a Twitter guy.

And so, I post on Twitter. I don't want to post on LinkedIn, but I have an agent that uh so it is a crown job. Wakes up every day at 9:00 a. m. this morning. And you can see it's logged in from my Twitter account. Like, this is authenticated as me. And so, it goes on my Twitter.

It's also authenticated as me on my LinkedIn. And it basically looks at the posts on my Twitter and it's like, "Hey, this is what I think you should post on on LinkedIn. " That's very cool. Yeah. Yeah, that makes a ton of sense. That's awesome.

And it can obviously like hydrate things and add extra context and really like transform the post properly as opposed to just like a default flow post on on X. Very excited to announce Lindy 3. 0. It's our most powerful Lindy and then it converts it to the the Lindy then converts it to LinkedIn.

It's like what's what I learned about B2B SAS convert I today we're transforming the power of agents to unlock productivity for everyone in your organization.

No, the real the the real clickbait for LinkedIn would be like uh you know employee three weeks ago I hired a software engineer who was working at 25 different companies. Here's what it taught me about building 3. 0 something like that. Anyway, uh well congrats on the launch. Anything else we should know?

Where can people sign up? Yeah, linda. ai. It's available today. Like I don't know about y I find it frustrating when like there's a lot of like vapor wear and AI. It's like this is coming one day. It's like no no no this is available right now. It works. It's incredible. Very What's the pricing model right now?

Don't worry about it. Consumption or subscription? It's consumption. It starts at $50 a month and it's it's a tenth of the price of a human. Okay. This is gonna be one of those things. Am I going to hear horror stories about people that like forgot to, you know, turn off their Lindy and they ran up a $10,000 bill?

Don't worry about it. It's the cost of inference is declining rapidly. It's all going to Everybody's going to make it. We're all going to be fine. Yeah. Yeah. We're going to distill the models. These aren't I mean, look, I you know, this has happened before and we're very generous with credits. People reach out to us.

They're like, "Can I get a refund? This happened? " And we're like, "Well, yeah, of course that. " Yeah, that makes sense. Um, cool. any other any other tools or or custom models or like RL environments that you're tracking or excited for progress to be made?

Uh you mentioned computer use, but what what else do you want uh the foundation labs to solve to kind of so that you're like a beneficiary riding the wave? Memory. Memory definitely. Memory I think is is the big gap. And I I think like there's a lot of attempts to solve this problem across the industry.

But I'm noticing this pattern where it's like people and companies like ours that are stuck and cannot change the weights of the models. They think of all of those like convoluted schemes to like hack the model to do something it doesn't when really it belongs to to the model.

And when you talk to the researchers at the labs and you ask them like ah like we're working on memory this is what we're doing. They tell you like don't don't do it like let us let us do this like we'll do it.

this belongs to the model and the model is going to do it a thousandx better than you do and I expect much more advanced memory schemes where it's like basically the model is going to be retrained to come out in like a year or something like six or 12 months.

What are you expecting from GPT5 just I'm expecting it to be smarter. Um, look, we feel sparks of a GI daily at this point. Like we had this moment last week like we basically accidentally built Lovable like because we gave a computer to Lindy like someone it was like 10 p. m.

at the office like Flo like hey if you ask me to to build a website she builds it and she deploys it it works and so like to me that was like a huge holy [ __ ] like like AGI Spark of AGI moment but like look it's still not as good as a human like sometimes it's just dumb.

It just makes dumb mistakes still today, you know. And so I'm expecting GPT5 to be like a step change and and to be a big step toward GI, but I still people don't realize what's happening. And I get it. People are tired of like all the AI hype and the noise.

And I tell them like if you're tired, man, like this is only the beginning. Like it's going to get a lot more noisy out there. Like it's it's gonna get pretty crazy. Yeah.

I mean, I I I'm just It should be exciting for everyone because Lindy uh today feels like the promise is is automate these tasks that are boring that a human does. The first 10 times they do it, it's interesting. The next 10,000 times they do it is very boring.

And so you just free people up to do these more like higher lever 0ero to one type activities that can then be automated and systemized over time. So um very very cool. Congrats on the launch. I'm sure, you know, people are going to be, you know, making bets on Lindy 4. 0. Is Lindy 4.

0 going to be able to get a college degree for me, you know, all that good stuff. Congratulations. Ironic is that is that these models are actually today able to get college degree so they can job as an intern, right? From the top to the bottom. So, yeah. Yeah. Awesome. Well, congratulations. We'll talk to you soon.

Great to see you. Thank you. Have a great day. Bye. You do. Up next, we have Dave from Rune Technologies coming in to the studio. Get that gong ready. Jordy Hayes and welcome to the