Weirdos wanted

This post is adapted from an all-company memo I sent the team

Hex Nation -

A ML model generate inferences, governed by the statistical distributions of its training data. This makes understanding what is “in distribution” for models an important concept for building with AI – because that’s what you can expect to get out of them.

And as much as these models can feel human-like, they can't have a creative or new thought that exists outside of their distribution. They are prisoners to their training data; just sophisticated stochastic parrots. And while they might do something interesting, they will never do something truly weird.

As AI automates more daily activities, more things will be done in distribution – increasingly reverting to the mean. And that's fine for a lot of tasks! There are processes where you want them done predictably, and we can hand these off to agents with increasing confidence. And what this suggests is that the role of humans will increasingly be focused on the things that live out-of-distribution – the things that aren’t reliable and repeatable and rote.

I believe this is the single most important thing to consider as we grow the team at Hex. We are not just looking for competence or charisma – rather for people think and act outside of the distribution.

In other words, we’re hunting for weirdness – the literal property of being unusual, different, or strange.

That might seem like a funny thing to look for - but it’s the weirdos that move the world forward. The nerds. The dorks. The tinkerers. The pirates. The people who will turn something on its head and look at it in a new way. The people who have the ideas that seem implausible - but come, eventually, to feel inevitable. The people who chafe against the strictures of what came before, and are impatient to invent the future.

These weirdos live outside of the distribution, and do the things that AI cannot.

And I believe this is why we've been so successful so far - Hex has a bunch of weirdos! I say that as the ultimate form of endearment and admiration; it's something I'm extremely proud of. Everything great about Hex has come from out-of-distribution thinking. Every original product idea, every creative solution to a customer problem, and every unique brand expression has come from a weird, and uniquely human, place.

We need to keep this going as we scale. That can be hard, because as organizations grow, they become more conservative, loss averse, and process-driven – and tend to hire people who fit and reinforce that mold. We must fight this with every fiber of our being! The weirdos are more important than ever, because the way to win in this new era is not just executing on defined playbooks - it's thinking from first principles, generating new ideas, and learning rapidly. The era of prizing predictability and consistency is over.

For all of us actively hiring onto our teams, we need to think critically and honestly about whether candidates are truly out of distribution. We must ask ourselves: what makes this person uniquely human? What is their alpha over an AI agent that can crank out in-distribution responses? Do they have a track record of breaking the mold? Bringing new perspectives? Do they have a side hobby? What rabbit holes have they gone down recently? Can you get out-of-distribution tokens from them in your interview? What makes them weird? 

I don't care what role you are in or hiring for, there is room for weirdness. We should always be looking for opportunities to bet on someone who is outside the distribution and can offer a lot of upside, vs. the "safe" candidate who can only offer predictability.

And for all of us who are already here at Hex, take this as a call to embrace your own weirdness. Don't revert to the mean. Don't censor your craziest ideas. Don't give in to the tyranny of the average. Let your freak flag fly. Ask the AI for help – but don't substitute that for your own original ideas. Be human.

In other words, live outside the distribution!

🖖💜,

Barry