Small matters in open tech

And other reflections from my chat with Jesse Pollak and Conner Ruhl on building open tech in crypto and AI

This morning I got to chat with Jesse Pollak of Base and Conner Ruhl of Stability AI — two builders and companies creating world-scale digital infrastructure and laying the tracks for future innovations. It was an awesome chat, so I thought I’d summarize my key takeaways. 

On a meta level, this conversation validated the idea that there are deep and important parallels between different frontier areas like crypto and AI — something we saw in our research on early stage founders as well. I’m quite confident you could map these same insights to biotech, space, or other frontier areas of tech, even though we don’t have as much personal experience with those (yet).

If you want to watch the whole conversation, you can view it here. It was recorded as part of 100 Builders, a 4 week cohort with 100 projects building open source in crypto and AI that we kicked off on Monday. As someone who has written about and worked on the value of community owned tech for years, I feel very lucky to be able to have these types of conversations with such smart and inspiring people. These notes won’t do it justice.

Small matters in open tech

One of the ‘aha’ moments was when Conner was talking about how important it is for Stability to keep their models like Stable Diffusion small enough that they don’t require supercomputers and GPUs that only a few companies have. Otherwise it doesn’t really matter that they’re open source because the cost of using them naturally leads to its own form of centralization. Jesse pointed out that you can think about the core design decision in Ethereum the same way, even relative to other chains – it needs to be small enough that anyone can theoretically run a node, otherwise it will also naturally get more centralized over time.

Much has been written about both of those topics, and both are nuanced. But something that’s clear is that the battle of open vs closed tech is much more of a spectrum than a binary. A model might be open source but keep the training data proprietary, an app built on a blockchain might have all of its data and smart contracts on a public ledger but have a proprietary app built on top. As this topic becomes more and more important (which it will) we’ll need cleaner mental models to view the gradient of open vs closed.

Now is the time to obsess about iteration cycles

Two things seem profoundly true about building in frontier tech at the moment:

• There is an explosion of technology aimed at making almost anything — in your product or in your workflow — better. You can outsource your coding workflow to an AI, or leverage some novel tech to build an app or feature that wasn’t possible yesterday.

• Because of the above, the ability to get lost especially in the search for product market fit is huge.

Conner and Jesse both talked about how if you take those things together, the answer on how to build is really just the fundamentals — ship fast and frequently, learn, and iterate. That’s something I’ve felt personally in crypto, but it’s clearly true across the board.

Once it's out there, it's out there

Fast cycle times are great, but it’s also clear that “move fast and break things” can’t be the motto of building open tech in a company like Stability or Coinbase. Conner talked about the need to bake your values into the tech when you release it, because once it’s in the hands of the community, you can’t take it back. And for Jesse and Base, when people put things onchain they are literally there forever. To find the complex questions about this stuff, start digging at the intersection of the need for fast iteration cycles and the fact that frontier tech has a Prometheus type effect. We’re getting fire, except this time it can be used to create rather than to incinerate.

The route to builder energy is to shorten the distance between "I have this idea" and "I built that idea"

For anyone building open tech, and for more and more companies in general, the ability to attract builders is life or death. We call it “builder energy” and think it’s so important that we’ve started to orient our whole company around it. Base and Stability have both been absolute rocketships from launch in terms of capturing and helping builder energy flow, so I asked them how they think about creating a magnet for it. What Conner said will stick with me, that the way to attract builder energy is to reduce the time between “I have this idea” and “I built that idea.” 

If you zoom out, it feels the shortening distance between idea and shipped idea describes the foundational paradigm shift we are undergoing as humans — increasingly, the average person will soon be able to create basically whatever they want. I’m personally very optimistic about the result, but there’s no doubt that how we build those capabilities will be one of the most important things we figure out as a society. There’s no doubt in my mind that our odds will be a lot better if we have more people like Jesse and Conner working on that, and thinking about the importance of open tech.