Firebase co-founder Andrew Lee returns to YC with AI agent platform Tasklet, now at $7M ARR

Speaker 2: And also has some big news today. We're gonna get to it.

Speaker 1: We'll get to that after our first interview with Andrew Lee from Tasklit. He's the founder and CEO. Andrew, welcome to the show. Thank you for the patience and joining us on such a busy day. How How is demo day going? How are you?

Speaker 4: I'm doing great. Thanks for having me. I wanna be upfront. I can neither confirm nor deny the SpaceX acquisition rumors.

Speaker 1: Oh, okay. Flying around. They're

Speaker 2: flying around demo day.

Speaker 1: Elon Elon's probably there at demo day, shaking hands, trying to scoop up, roll up the entire batch maybe. Who knows? He's got stock to do it.

Speaker 4: No comment. Demo demo day is great. We're we're we're having a good time. There's a thousand people here. There's 200 companies. Wow. It's it's a good time.

Speaker 2: Amazing. Did you did you pitch already? Or is that ahead of you still?

Speaker 4: Not not yet. We're in we're in group four, which is gonna be this afternoon. And it's actually my cofounder who's gonna be up on up on stage giving the

Speaker 1: talk. Cool.

Speaker 4: I'll just be the pretty face over on the side.

Speaker 1: Okay. Well Perfect. Give us some backstory on yourself and tell us about the business.

Speaker 4: Yeah. So as background, I was the thing I'm probably best known for is I was one of the founders of Firebase.

Speaker 1: Oh, yeah.

Speaker 4: So this is my second time through YC. I did the summer two thousand eleven batch. Yeah. We sold that company to Google in Yeah. '20 I was there for about three years. This this company has put a very interesting story. We actually started in 2020 as a better Gmail. Like we tried to build a new email experience. Raised a bunch of money, hired a team

Speaker 2: So you quit Google to try to to to try to eat Gmail alive.

Speaker 7: Yeah. I

Speaker 4: I had already left Google a couple years before but I saw them shut down inbox.

Speaker 1: I was

Speaker 4: like, man,

Speaker 8: what are these This is

Speaker 4: a huge huge product. Like, are these guys doing? We should do a better job. Yeah. This is a terrible idea. Don't do it. Don't invest in it. But just as we were getting ready to shut that thing down, LLMs got good. And we're like, holy crap, we can take all of this data that you have and we can feed into a language model and we can draft emails for you. We can do interesting types of search and we can do categorizations. So we started kind of pulling that string. And around about this time last year, we had this really good agent inside our email client and we had a bunch of users being like, Hey, I love your product, but could you just get rid of all this inbox crap? Like the UI is getting in the way. And so we spun it out and we had a that's a totally new product, totally new code base and this is Tasklet. And we launched that in October and that's gone super, super well. And that's why we're in YC. And that's gone at the beginning of the year, we were at about a third of a million of run rate. We're at a $7,000,000 run rate now. So it's like a

Speaker 2: nice Woah.

Speaker 4: Nice fast ramp. And

Speaker 2: Is that the bar now? Like is have have is there you know, are you meeting other companies in the batch that are that have been able to ramp that that quickly?

Speaker 4: I I think we have the highest revenue of the batch.

Speaker 2: There we go. Let's go. There we go. Well, why don't we go hit the Gong? Oh, yes. And then then we can Boom. That's for you. Love that. So feels like day. So we were we were looking at at at our little summary here. AI agents that connect across work tools run twenty four seven and take ownership of recurring workflows.

Speaker 1: Does big tech hate you? Like, he is I feel like the the whole point is walled gardens. That what that's what creates value accrual. That's what keeps the value accruing to the work tools that, you know, big company CEOs love and extract a lot of value from. And, you're coming in here and digging a hole under the wall of the garden and throwing a ladder over the top and digging a hole right through the the walled garden. Absolutely. How do you, like, maintain those integrations if if if the actual platform is maybe, like, hostile? We've seen reporting about this where different companies have tried to sort of integrate across and then they've wound up with sharp elbows. I think a lot of them have figured it out. But, like, how do you think about working copacetically with the the work tools that you integrate with?

Speaker 4: Totally. So so AI has totally changed the game. You know, two years ago, if you want to integrate with a bunch of stuff, you had to hand code a bunch of integrations. And there were companies like AnyDan and Zapier that just did that. And that was their moat. And our secret sauce is that we can dynamically generate the integrations with AI when you need them. So we have a bunch of canned integrations. Some of them we've hand coded and made them really, really nice. But even if you ask for an integration we don't have, the AI can generate one for you. And this is super powerful, not just because it lets you go outside the walled garden, but because it lets you connect to things that not be public at all. So if you're a company, a lot of times you have internal APIs and you're like, hey, want have some integration that hooks up to Notion and HubSpot and Gmail, but then also my random bespoke internal API. And you can't do that with any of the products from big tech. You can't do that with Yeah. You know, Claude or or OpenAI. Yeah. But our product yeah. Sure. You can integrate, you know, via API with this thing, with MCP, with this thing, with a a pre canned connector with this other thing, and they can all work together.

Speaker 1: Mhmm. What's the state of search these days? I feel like we've seen incredible progress in AI and agents. And then when I have the same frustration that you had when you were thinking about working on email, I'll type in some keyword, you know, like y c. And I will get some cookie, some string that's stuffed in a receipt from some coffee shop, but it has nothing to do with Y Combinator. The context isn't there. Apple intelligence rolling out. It takes, like, five days just to index everything into some sort of rag database. Like, why are we falling behind there? Is that a key is is that key to your strategy, or is it something that it's a it's a problem for someone else to work on?

Speaker 4: So I think agents, again, have totally changed the game here. So in shortwave, our email client. Yeah. The way we did search was the vector database approach, right? Embedded all your emails. We stuck them in a vector database. We had this interesting search stack and the goal there was to have like the search results be good.

Speaker 1: Yeah.

Speaker 4: And it turns out with, if you have a really smart model and you have an agent, you don't actually care that much if the search results are good as long as you can run lots of searches. So what we found in Tasklet is we don't need to like ingest all your email and index it. We can hook up to the regular APIs and just run a whole bunch of queries in parallel, look at the results, adapt, iterate.

Speaker 1: And just keep refining.

Speaker 4: And so the results are are are almost as good without any infrastructure. So it's a lot cheaper. You can connect to more stuff. So Yep. I think if you find use Tasklet, you can kind of hook it up to anything. Yeah. And you'll get surprisingly good search results without any special back end

Speaker 1: support. I imagine that that's sort of slow right now. And for a company like Google to roll that out to a billion users, you know, for free immediately, that's going to be really expensive. So that's still a ways out. They need to actually chop through that. But we could expect that search will get faster and better over the next couple of years.

Speaker 4: Yeah. I think I think that is the big drawback. And like as an example

Speaker 1: Yeah.

Speaker 4: For prep for Demo Day, I wanted to make sure that I chatted with the folks that were here that had reached out to me. I've had people reaching out for a period of months. So I took Fable five back when that was still a thing. Sure. And I pointed that at my Gmail and I said, Go find everyone who's who has cold emailed me Mhmm. In the last ten months who might be at demo day and like give me a spreadsheet with those folks. And it spent like and and and stack rank them. Right? Like I want to know who should I

Speaker 5: talk to first and

Speaker 4: so on. And it spent like half an hour. It spent like $60 running this thing. Right? But the end of it, I had this spreadsheet of like a 100 plus folks that was like carefully researched and like ordered and like it found every like there was no one I could think of that it that it didn't find in that process. So So if you're willing to spend $60, if it's worth it to you, like, great.

Speaker 2: I love it. Are you last question. Are you capital constrained? Like, I'm assuming you have a bunch of offers. Maybe you already finished your fundraise. But why even raise? It feels like you guys seem to be making I I would imagine with that kind of revenue ramp, it'd be hard to not be making money right now.

Speaker 4: We we are we are not capital constrained. We actually already closed a $20,000,000 round in April with There you with USB and Lightspeed and some other folks. But they're you know, we're competing directly with Anthropic and OpenAI. They've obviously raised a lot more money and I think we can put it to use. So we are we are raising more money now. We're probably a bunch

Speaker 2: of reps. Just I like just saying the actual reality because there's a lot of founders that would come on and be like, well, actually like Yeah. We do something that like, you know, and it's like, no. Everyone is competing all the time and it's better to just accept it and play the game. So

Speaker 4: 100%. And I think Corgi set a new bar with, you know, three weeks from from was it series b to c or whatever they did there. So, you know, why wait?

Speaker 2: Three days. Do it in three days.

Speaker 1: Three days. I love it. Awesome. Thank you for coming on.

Speaker 2: Yeah. Great to meet you, Andrew.

Speaker 1: Congratulations. Congrats

Speaker 2: on all

Speaker 1: the We'll talk to you soon.

Speaker 6: Talk to you soon.

Speaker 5: Thanks for having me.

Speaker 1: Have a good one. Cheers. Me tell you all about the New York Stock Exchange. Wanna change the world, raise capital at the New York Stock Exchange. Our next guest will be sitting in that same seat slotting in in just a second. We have

Speaker 2: Stamatios will be working

Speaker 1: on Eden

Speaker 2: Robotics. The microphone.

Speaker 1: Oh, yes. Yes. Yes.

Speaker 2: Good Alexander says, I pay for Starlink. It's a great product. Cursor is also a great product. Codex is a great product. And Claude with Fable is ridiculous product. That's the main difference between now and 1999. All this stuff is awesome.

Speaker 1: So total white pill, then read the follow-up.

Speaker 2: I don't think you can really compare this to 1999. I think it's actually unprecedented. If AI doesn't work, the global debt bomb explodes, and there's a great depression.

Speaker 1: The

Speaker 2: 1999 So did Edson say popping. Things were generally fine.

Speaker 1: Yeah. So it is higher stakes, but it's more impactful. So he's he's I mean, it's a good point to not try and draw too much comparison to 1999 just because it's it's it's technology. It's a technology driven boom. It is very, very different in terms of the the financial structure and also the the impact, the revenues, the rollout, the opportunity.