Finding Product-Market Fit and Riding the Wave

When I went through YC (many moons ago), I still remember hearing stories about startups with incredible traction and growth rates. The YC partners would use them as examples of companies that had found the elusive "product-market fit" and would advise cohorts of startups to follow in their footsteps. One of the companies was Teespring - at the time, it was the fastest-growing e-commerce company ever.

Which is why I was excited to talk to Evan Stites-Clayton, the co-founder and CTO of Teespring, who now advises his own cohorts of startups as a partner at HF0. HF0 is a 12-week residency program for technical founders, with a recent focus on AI. They've been profiled in the New York Times, and have already had some impressive companies graduate from the program. They also have incredibly generous terms - applicants give up 3% of their companies for $500K in investment (in contrast to YCombinator's 7% take).

Evan and I talked about his early days at Teespring, his transition into AI, and some of the projects he's most excited about - including companies like Krea, iKHOR, and Recursal.

Three key takeaways

Finding distribution channels is a key part of hypergrowth. That said, it's unclear exactly how AI will impact distribution - while it offers a significant technology shift, how can companies use it to get in front of more users?

I think that's also a part of what makes a lot of really great companies tick is that not only do they have the determination and grit and they build a product that people actually want, but they're also riding some kind of wave of, like for us it was Facebook.

It was Facebook as a distribution channel, whether it was Facebook groups or then later Facebook ads. But that was the thing that led us to be able to get to that hypergrowth. and I think a lot of companies right now are trying to ride like an AI wave, but it's also important to ask yourself, well what is the AI wave?

With Facebook, it was actually pretty obvious - it was this new distribution channel. It was a new way to get into people's feeds. It was a new way to get, an ad in front of someone basically. Now with AI, what is that?

Lifelike assistants are an exciting (and underdeveloped) aspect of AI. One of Evan's big interests in AI is digital assistants - he believes it's a major upgrade for consumers to have different personas and friends available 24/7.

What I'm most fascinated by is ... a digital friend or a digital assistant that's truly in my life in a way that a human would otherwise be. That's where I think massive value gets unlocked because you enter this world where everybody is able to have all of these different roles in their life. The coach or the helper or whatever it is.

But I don't think we're there yet. I think there are a lot of fundamental advances that still need to happen. And it's sort of a UX paradigm of asking ourselves, what is the UX of human interaction that actually creates compelling exchanges? A lot of people think you just can't do that with AI - and I don't agree with that.

AI startups need to lean into their agility. This is a sentiment echoed by Andrew Lee of Shortwave - startups have a pretty big advantage in how quickly they can move. Even though incumbents will eventually try to copy them, agile startups will already be on the next frontier of AI features.

You have the advantage of having none of that baggage and just being able to completely run in a new direction, a new paradigm. I honestly think the best thing for most early stage founders to do when it comes to thinking about "What if Google builds this? What if Adobe builds this?" is to just ignore those fears and just continue to forge ahead and continue to get users, and continue to listen to the users and continue to build the thing that resonates.

...

So ideally you're a position where, let's say you're Krea, you build your AI image generation suite, and then Adobe comes out with its stuff. And by the time Adobe's coming out with its stuff, you're already moving on to the next thing that you're about to release. And so you continue to stay ahead. You continue to stay one step ahead, and then Adobe's gonna release all that stuff they're gonna have to support that.

And so you need to keep forging ahead and building more new stuff that's going to continue to keep you relevant. And I think it's very possible for startups to do that because they have less users. They have scrappier dev teams usually, and there's just so much less overhead to like shipping a feature If you are a ten person engineering team than if you are a Google or an Adobe.

And three things I learned for the first time:

* By 2015, Teespring was doing a million dollars a day in sales.

* In contrast to the Transformer architecture that's dominated modern generative AI models, there's an alternative approach called RWKV enabling other types of AI - with different pros and cons.

* HF0 originally started as a general-purpose program that changed locations with every cohort. That changed when the pandemic struck (and when Evan hunkered down in their house for months).

Artificial Ignorance is reader-supported. If you found this interesting or insightful, consider becoming a free or paid subscriber.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.ignorance.ai/subscribe