Last Saturday I spent the day at Inspire9 in Richmond for the Claude Impact Lab, the first one Anthropic and the Claude Community have run in the Southern Hemisphere. The brief was good. A roomful of builders, one day, and the City of Melbourne’s open data: permitting records, parking systems, public safety, council meeting transcripts. Most of it sits locked in clunky portals that the residents who need it can’t use. Point your skills at something that matters. Go.

The room was extraordinary. The format wasn’t built for it. Here’s what I mean.

The room was the win

People were vibrating with excitement. Not the polite conference hum — actual leaning-forward energy. I asked seven different people what they were using AI for and got seven completely different answers, every one of them interesting.

A real cross-section came through the door: more women, more people of colour, younger than you usually see at a Melbourne tech event. That doesn’t happen by accident. Someone worked at it, and it showed. The registration app was well designed. The datasets were solid. The mentors were sharp. Anthropic put real support behind it.

The young woman beside me was a bioinformatician at MCRI, working on genomics, used Claude for everything. The guy at the entry coached basketball and was using Claude to narrate post-game footage for player development. A Pakistani computer engineer making the event video. A recent-grad marketer running her own AI-assisted consultancy, breathlessly excited. The room was the people. Anthropic and the Claude Community brought the people. Credit where it’s due.

The format spent them

Eleven people, one deliverable, a clock. There’s only one way that goes.

We circled for ninety minutes — introductions, ideas, the two most confident voices easing to the front. I lit the fuse myself: “you just need somebody to say, okay, this is what we’re gonna do, and then do it.” That handed the wheel to whoever was quickest to grab it, and in that room that was never going to be the quiet ones. “Should we just start running, let’s just put a bunch of tickets up there.” “You happy to delegate, or happy for the design team to come up with a feature set?” Engineering lane, design lane, split by the two men doing the splitting. First artefact: a Jira board. We still hadn’t said the word Claude.

This is the dynamic the format is built to produce. Eleven people is a committee. A committee under time pressure collapses into whoever takes the wheel. In a tech-coded room, that’s the people whose whole life has trained them to take it. In a room organised specifically for diversity, the format hands the room to the loudest voices.

Some attendees didn’t have a single good interaction all day. Quietly. Brilliant people, completely excited and empowered by these tools, who walked into a room organised around its loudest voices and never got heard. They came in capable, excited, and left flattened. In a room assembled specifically to be diverse, that is not a small thing.

This isn’t a Rye Smith problem. Rye built a venue, 70 attendees chosen from 430 signups, live-coded the check-in app that morning, ran the Discord, vetted the mentors, brought the City of Melbourne into the room. He delivered the event he was asked to deliver. And there’s no global playbook handed down for these — each Lab is shaped locally, by the people running it, learning as they go. That’s the hopeful part: the shape isn’t fixed. Team sizes, structure, the whole arc of the day are choices, and choices can be made differently next time. The shape is the question — and it’s one the organisers get to answer.

The salon they could have run

If the gold is the conversation, design for the conversation.

Lightning talks all day, short and sharp: “here’s the maddest thing I’ve done with AI.” Who are you, why are you here, what will you build to change the world? A communal wiki the room fills in together. Record the conversations, hand the transcript to Claude, have it sort the room live — birds of a feather in this corner, birds apart in that one. Small groups of three or four. Fifteen minutes. Rotate. “Turn to the person next to you and tell them your best prompt.” That’s the event.

There’s a coordination problem buried in it. We’re physically in the room. A room full of people is exactly what’s good at solving coordination problems — if it’s structured to do that work. The current shape isn’t.

The MCRI bioinformatician beside me had a hundred questions she wanted to ask the room, and a hundred bits of wisdom to share. How do you deal with shadow AI in a privacy-restricted environment. How do you convince your boss to buy Claude Code. How do you work around institutional restrictions on AI tools. Those are the actual questions the Claude-curious workforce is asking. None of them were the day’s questions, because the day was a build-a-product hackathon and her questions are culture-change questions. The technology is the easiest part of this. We don’t need a competition to prove what the tools can do; we need a forum to figure out how to live with them.

The question nobody hacked on

Everyone in that room will be accelerated by this technology. We’re the early, the curious, the already-capable. Even the ones who got flattened at my table will be fine — they walked in capable and excited, and that won’t change. The Impact Lab was rightly about civic data and city services. But the deeper civic question was sitting right there in the room, and the format wasn’t built to see it.

The nine people who got steamrolled worked to be heard in the ways they could, in the gaps between the loud voices, like people do. But they shouldn’t have had to. A format built for them would have rewarded them; this one rewarded volume. The woman beside me, hungry, sharp, full of the right questions, never got the room. And here’s the tell: eleven capable people, the most powerful thinking tool ever built, open on every laptop, and we didn’t reach for it. We didn’t mention Claude. We reached for a Jira board and an old pattern. We had something that might have helped us hold the room differently, and instead we reproduced exactly what we always do. Those were the lucky ones — they at least made it into the room. If this is what we do to the sharp people at the table, with the tool open in front of us, ask what happens to everyone who never gets through the door.

That question — the people who aren’t in the room — got asked from the floor. Anthropic’s Martine Nilsen, APAC consumer programs, running the day from the stage, took it. Her answer was honest. “We’re at the very beginning of something… early adopters are always gonna skew a certain way.” She didn’t dodge it. She named the strategy as real but unlaunched, and called the hard part by its name: “how do we open up and create the spaces in an inclusive way, an equitable way, to gather more voices… and bring the people, like you said, to the table?”

She’s right. That is the challenge. It’s also the work the format spent.

And yes — everyone has agency. I’m not arguing anyone’s a victim. But agency isn’t evenly distributed, and a format that rewards whoever grabs the wheel fastest hands the future to the people who already had it. We don’t have to rebuild the world we’ve already got. We get to choose.

I’ve spent years watching this technology get taken up by real people, and it never lands evenly. What works is slower and more human: give people a chance to talk about it, to think about it, to ask how it touches their own work. Show and tell. Maybe a hackathon is how you get the tech people in the door. Fine. But it’s the wrong shape for the thing that actually matters, which is everyone else.

If that’s the lab Anthropic wants to run, fine. But then call it that. “Impact Lab” makes a claim the program’s shape doesn’t cash.

The hardest question about AI is not how it works for the people in that room. They’re fine. The hardest question is how it works for the people who weren’t — the ones this wave leaves behind. It got asked. It got acknowledged from the stage. Then the day went back to building apps for council. That’s the gap that matters. The next one should be built to close it.

Thanks where they’re due

Rye Smith and the Claude Community got this off the ground. First one in the Southern Hemisphere. The hard part — the room — they nailed. The format was a local call, not a fixture — which means the next one can be a different shape. I’ll be there. I hope it is.