Mastermind · February 3, 2026
Hackathon week is the real test
The week ran two hackathons in parallel, one at Johns Hopkins and one at Stanford, and the honest takeaway is that we have distribution and we don't have reliability. Those are not the same problem.
I came into Monday's mastermind still riding the signal from the events. New users who found us through the hackathon funnels. New partners asking what it would take to plug into our event layer. New intros from sponsors who'd been watching us but hadn't moved before. From the outside, the weekend looked like a win. From the inside, I was watching edge cases fire in real time and keeping the platform up by being in Slack at 2am.
Xisen pushed on the part I'd been softening. Distribution is always the easier half. You can manufacture distribution with events, with content, with relationships, with partnerships. The harder half is whether the product holds up once distribution works. Most teams underinvest in reliability because reliability is expensive, invisible, and only pays off during an event. Every hackathon we run is effectively a paid QA pass on our platform. The question is whether we treat it that way.
The pattern I kept seeing at the events was the same one. A submission flow that works at 10 teams breaks at 50. A judge interface that works for one judge breaks when six judges are hitting it at once. A sponsor handoff that works when I'm on the floor breaks when I'm not. None of those are deep bugs. All of them compound into a bad experience for an organizer who has three hundred participants to wrangle and no tolerance for us not working.
So the shift this week is treating every hackathon as the real test of the platform, not the marketing moment. The operational spine has to hold without me in the room. That means the onboarding flow, the submission flow, the judging flow, and the organizer admin flow each have to survive 48 hours of event-level load without requiring a founder to babysit. Until that's true, we don't actually have a platform, we have a service wrapped around one.
The competitive landscape conversation also sharpened on the call. There's a US competitor a couple years ahead of us that's running events with the largest AI labs and running up real revenue off those contracts. Watching them, you see the same pattern every platform company repeats: they invested early in a social layer and event discovery, and they got a flywheel that's hard to unwind. We can't beat them head-on in the labs they already own. What we can do is differentiate on the organizer side and on the markets they haven't prioritized yet, specifically ecosystem partners in Asia and builder communities they don't touch.
What I want to avoid is the trap of collecting partnership logos. Every conversation that doesn't produce a concrete distribution test is a warm thread that costs attention. The partnerships I'm pushing forward this week are the ones with a real next-step deliverable, either a co-run event, a shared funnel, or a signed pilot. Everything else goes to the warm list and doesn't get weekly calendar time.
Xisen is fighting the same problem from the research side. He's tightening the network-of-agents paper and pushing a better testing cadence into the Pulse team. The shape of the issue is identical: you cannot run a real research or product agenda when your team can't reproduce its own results. Testing discipline on his side is roughly what reliability discipline is on mine. Neither sounds exciting. Both are the difference between a side project and a business.
The part of the call I want to write down because I'll want to remember it: when you run the events yourself, you see your own product the way a stranger does. A stranger doesn't care that the feature is new, they care whether it works under the conditions they care about. This weekend was the first time I understood in my body, not just on a dashboard, which of our features we actually ship with confidence and which we ship and then hope.
The check I'm making on myself: by the next event, the organizer admin flow has to be good enough that I don't have to be on the floor. If that isn't true by the next mastermind, I'm the bottleneck, and the writeup about how distribution is working is a distraction from why the platform isn't.