Dan Shipper spins out Good Start Labs from Every — AI models that learn through gameplay, raises $3.6M

step up. I mean, the the story of Netswuite was uh was an acquisition. Next up, our guests uh nominative determinism is absolutely insane. We have Dan Shipper who is shipping other product. Let's bring him in. Something the shipper himself. What are you shipping? Did you do people do people ask you about that?

Did you say like does he ever ship? He does. I have gotten that. I have gotten that. Um and I don't know if my ancestors were nautical or not, but um certainly it works for the startup world. So uh yeah, shipped a couple new things today. What' you ship? Give us a news.

Okay, first first big thing is um we shipped a new uh incubation. We incubated a company. Uh we spun it out. Uh it raised $3. 6 million from General Catalyst and Anovia. I got the gong, the first ever gong for me. I'm excited. Um and it's called Good Start Labs.

It's run by Alex Duffy, who was previously our head of AI training. And the idea of good start is it teaches AI models to play games. Um and by playing games uh uh they're able to generate high quality reinforce reinforcement learn data and pre-training data that helps make models better.

Um it started as a product a project inside of every that we launched as AI diplomacy where we taught um AI models to battle each other for world domination. Uh it went super viral. It had like 50,000 people watch it on Twitch. Um, and it taught us a lot about um, how models work.

So, for example, Claude never lies in diplomacy. What the whole point of diplomacy is to backstab your opponents and Claude never lies. Um, it's really interesting and Alex, you know, the response to that was was really incredible.

Alex has a has a vision for how not just diplomacy but all sorts of games can be used as dynamic environments to um help models learn to be smarter in all sorts of different ways from long horizon task planning to writing to humor to um uh training them to not be not be as deceptive all that kind of stuff.

I'm mostly hearing AI that lies though so we can win diplomacy sounds sounds but no I I I completely take your point. Claude has to launch liar mode.

I I mean, yeah, there's so many absolutely right like there's so many examples we could go into about alignment by default like resulting in like not what you actually want, a hallucination sometimes a feature. There's a million different uh areas there. Do you have any question?

What how how concentrated will the customer base be for this? Is this the kind of thing you're going to work with hopefully all the big labs really quickly? Um what does that look like?

Yeah, I mean I think I think the the idea for the company is um is being a vendor to all the big labs um and anyone who is who wants to be a big lab. Um so yes, the the the customer base at least initially is going to be going to be pretty concentrated. Uh rebutt Will Manitis.

He said he's calling the top on incubations. You're still doing them. Put him in the truth zone. Break it down for me. Why are incubations still a good idea?

This feels this feels like, you know, uh, to to come to Dan's defense on this one, having an employee at your company who's proven themselves internally and discovers a good idea and then decides to launch a spinout company is very different than like, I have an idea, let me go find a random person, which is really hard, right?

Because you have this adverse selection bias. Yeah, I I agree. But we also do that and it works. Um and so so this is our second venturebacked incubation.

Um collectively, so uh Good Start Labs and our first one was called Lex and collectively they've raised like $6 million which is um more than like about more than three times or four times what we've raised for every itself. Yeah.

Um and we also run um four apps internally to every many of which are ideas that I or someone else inside of every has come up with and then we've recruited someone from our audience to take it and lead it and that works really well too.

So we've got four apps in market um and we grew collectively for every for you know all the subscriptions we sell we grew MR 50% in Q3 um so it's just growing super fast.

So you can really you actually can do this and I think you can uniquely do it in AI because you can um uh you can have these products be run by one person. Yeah. Which is really hard to uh help me uh synthesize this.

There's like this general mood on the timeline today that like a lot of the foundation model companies might be like plateauing and so they're going into other erotica play. No no it's not that that's not my question. My my my question is every we're going to do every I mean you're already a writing platform.

Why not spice it up a little bit? No, my my my spicy mode. There was a moment where the mood was uh like don't build a new company because you're going to get steamrololled by AGI. It's just going to be able to do exactly what you do but better.

And so like it's super risky to build anything that's in front of the steamroller that is the the big model companies.

Now uh we've seen that a lot of companies have popped up and done very well in you know other areas but now it also feels like the the the foundation model companies might be going after social media going after erotic content or adult content.

So is it is it more or less dangerous to build a startup that is in the wake of the of the AI uh like you know new hyperscalers? So I mean I think of every if you're a sapioexual we're already doing erotica for you. Um, so that's that's one thing. Got to get that out of the way. Yeah.

Um, but you know, when we first incubated Lex, that was like in 2022, 2021ish, and that was right in the in the very beginning. It was right before Chachi came out. And so it was in the very beginning of, you know, rapper anti-rapper hype. And that was a dumb take then, and I think it's a dumb take now.

Um, uh, incumbents always manage to things up. and um and and need to also uh you know if you're if you're a PM at OpenAI and they're super talented, they're doing amazing stuff and they also have to look for opportunities that look like they're going to be a billion or 10 billion or $50 billion in value.

And um there's just many many many different corners of the market that they just can't see. And I think and and and founders are incredible at finding those corners of the market.

And because of what these foundation model companies are building like OpenAI, uh you founders have never been in a better position to serve the the the needs that they find. And so I think it's I think it's the best time in the world to be starting companies.

And I think doing it in an incubation model like the way that we're doing it is kind of it's like one of those things that shouldn't work, but it's sort of uniquely interesting because we have distribution.

So in a uh in a world where anything can be vibe coded, we can give founders a like a stamp of approval like hey this thing is like actually um good. It's going to be supported. Um and there's a lot of interesting things you can do if you have a network of um apps that are all in the same ecosystem.

So uh I I'm I'm super excited about it and I you know I think every time we've called oh it's the end of this startup or this startup because OpenAI is doing it that's never happened. Yeah, makes sense. Uh what's your take on the regulatory debate and anthropics recent moves? A lot of people are pissed off about it.

From Keith Rboy to David Sachs to Mark Andre. I Okay, I I I agree to some extent.

Like I actually looking at what um you know the kinds of things that Daario says for example, I'm like how can you be building this stuff and then go on the news and be like it's going to replace all white collar workers in the next 18 months. I can't remember the exact quote, but like I just I don't like that.

And also, I think Sachs is wrong. Like, you're watching nerds who have a tremendous amount of power and responsibility grapple with that publicly. And I think they are being earnest. I think they're doing it in a sort of ham-handed way.

And I don't, you know, I think that they could they could clean that up and they should get smarter about it, but I don't think it's some like sophisticated regulatory capture scheme. And I imagine Sax is sort of telling telling on himself there.

like that's what he might do, but I don't think that's what like anthropic people are doing. Yeah. Genuine believers. Uh well, thank you so much for stopping by. Congratulations on the new on all the progress. Yeah, I was going to ask I I did want to push you. We do have somebody in the reream waiting.

We have one person in the waiting room, but uh what was your immediate rea like where where do you stand on the debate uh on the uh adult kind of content entertainment debate with uh with Chad GBT? I think anything they do, they're damned if they do, damned if they don't.

Um, if they allow it, they're damned because people are like, you know, you're going to turn this into a waifu platform and if they don't allow it, then it's a freedom of speech issue.

Seems like they're they're trying to be um thoughtful about it and uh restricting it to uh over over 18 people and not doing it by default, like only doing it if you if you select it. That that seems like a reasonable first pass at it.

I'm sure there are I'm sure there are so so open that open's mission is to ensure that AGI benefits all of humanity. Is it Do you think that it's hard to argue that that uh adult content porn bots benefit humanity? I'm sure that somebody can make a justification for that.

Do you think their justification is like this is a necessary step to generate the revenue that gets us to the scale that we need in order to deliver the ultimate goal?

I mean, I think that the caricature of, you know, uh, adult porn bots, uh, is obviously bad and and and there are, uh, there are places that you can take this that are obviously bad for people. And also, people choose to engage with adult content.

And I, you know, it's it's hard if you're trying to build something for everyone, which is what they're trying to do, to completely say, hey, like that's off limits.

Like that's a that's a particular moral stance that I think you can have smart defaults, but I also think it's within bounds for them to say um in certain circumstances for certain kinds of people, this is what they want and and they should be allowed to do the thing they want to do. Yeah.

It just feels like it feels like YouTube did YouTube has never gotten a ton of flak for not allowing adult content on the platform and being like YouTube's anti-free speech. YouTube has gotten a lot of your anti-free speech criticism, but never because there's not enough adult content on the platform.

It's always been because of depp prioritizing particular political thinking or something else that is like more in the wheelhouse of free speech. Everyone, at least in the tech zeitgeist, has kind of been like, "Yeah, it's fine that YouTube doesn't have adult content. That's their choice.

" And they just chose not to do it, I guess. I don't know. I'm sure the debate will continue to rage as it always does and we will continue to monitor it and thank you for coming on the show. Congratulations on the launch. Excited to see it play out. We'll talk to you soon. Have a good one.

Before we bring in our next guest, let me tell you about public. com investing