Inject the AI straight into my veins.

There’s a huge amount you can learn about becoming AI-native on your own. The labs’ white papers, blogs like Simon Willison’s, the better parts of AI-pilled Twitter. I’ve read a pile of it, and a lot of it is excellent. But at some point you hit a ceiling. Because reading about how someone works isn’t the same as watching them do it and being able to ask why, and how this applies to your specific circumstance if at all.

So as we’ve tried to push our whole company through an AI transformation, we went looking for the next best thing. Someone who’s actually deep in the weeds, possibly all the way into AI psychosis, and putting them in front of our team showing what’s really possible. In human form, in the room, taking questions.

That’s what we started doing. We call it a speaker swap.

Our first one

We ran the first swap in March, bringing in Jon Williams, co-founder of Pyn, a guy who stared into the AI abyss and climbed in.¹ Much of what he shared you can read on his blog. But once again it’s one thing to work through a stack of someone’s posts, and another to have them sit in your room and tell you which ideas actually mattered most. He could put the weight on the two or three things that moved the needle for Pyn, out of everything he’s written. We could ask about the exact problems we were stuck on, and half the time he’d already solved them. The generic stuff in the posts is the easy part. The weighting was what we couldn’t get from reading.

As an example he runs twenty-odd sessions in Claude at a time, one-shotting his prompts to get to an outcome, and deleting whatever doesn’t work so he can start again.² It sounds wasteful, and at first I was skeptical it could produce anything meaningful, but for Jon it’s a ladder that’s consistently pushed Pyn forward.

And he wasn’t on a stage. There was nothing performative about it, because it was just him and our twenty-odd person product, design and engineering team, talking builder to builder. He had no reason to overstate anything or bullshit about what’s possible. The best part was watching him share his screen and walk us through the actual code he’d written, the products he’d shipped, and the workflows he’d wired up, in enough detail that we could go and try it ourselves.

And the team walked away pumped at the possibilities, ready to try it all out for themselves.

What it is

The idea is simple. We bring people into Great Question who have a different relationship with AI than we do. Sometimes they’re miles ahead, occasionally they’re a step behind, sometimes they’re at exactly our stage, and you often can’t tell which stage they’re at until they’re in the room. Either way they’re showing real workflows, real tools, real wins and real dead ends, live, with time for the team to dig in. It gives us a human version of someone going through this same transformation, which makes it much easier to work out where we actually sit on the journey.

Then we return the favour. Someone from our team goes and gives the same kind of talk at another company, fields their questions, and that company sends someone back to us. It’s a swap³, not a guest lecture.

Why we do it

You can’t become AI-native from reading blog posts, the same way you can’t learn to cook or surf from reading books about it. You need to watch someone do it, ask them questions, and have them show you what it actually means in practice.

That’s doubly true when the state of the art moves every week. What was cutting edge this time last year is old news now, and a fair bit of what was cutting edge last month is too.

A swap also lets us cut through the crap. Plenty of people out there have good reasons to sell you their book, and plenty more are doing things that just don’t apply to a company like ours. Getting them in a room where we can interrogate it is the fastest way to tell the snake oil from the real thing, and to get to the real thing quickly.

What we considered instead

The obvious alternative is to hire it in. Bring on more applied AI engineers with a broader range of backgrounds than we have on the team, folks who already work this way. First of all that takes months, second you don’t really know what you’re getting until they start, and finally it’s one person with one point of view. Hire ten and you’d get more coverage, but now it’s enormously expensive and you’ve still boxed yourself into ten people who mostly think alike.

The swaps gave us a much broader spectrum, fast. We’ve had founders, VPs of engineering, UX researchers who moonlight as product builders, hobbyists, and engineering managers come through. One of them had vibe-coded something that looked a lot like a piece of our own product, which told us more about where the competitive threat comes from than any teardown would. No single hire can give you that range, let alone that quickly.

How to run one

Keep it light. The speaker spends a few minutes on who they are and why you should listen, then gets straight into a real workflow. “Here’s something I did this week that’s a bit wild.” Then questions, which is where most of the value lands.

Invite all of your team, not just engineering, so that everyone can have those “feel the AGI” moments and figure out how the new technology applies to their work. Ideally record it, but be willing to turn that off if you think it’ll lead to significantly more candor in the learnings.

The hardest part is finding the speakers, and the thing that’s worked best is just being loud that we do this and asking around our network for suitable folks.

Since I started talking about it on LinkedIn, people have reached out wanting to send someone, or putting their own hand up to come and talk. I’d love more of that, because everyone in our position is trying to crack the same thing at the same time, and most of us are doing it too quietly.

So if you can’t yet build AI into your company’s DNA, borrow someone else’s. Get the people a few steps ahead into a room with your team, send your own people out to do the same, and do it again next month, because by then the frontier will have moved.

New posts land weekly. Subscribe at clubned.co and they’ll come to you.