Mastermind · February 14, 2026
Two features, done better than anyone
The line that crystallized the week's call was Chinat's: the product that wins an early market is usually the one that does one thing so well nobody questions why you built it.
I'd come in with a Pulse roadmap that had become too generous with itself. Automated outreach. Presence across a few public surfaces. Meeting notes. Live transcription. Visual memory. Persistent context. Agent-to-agent communication. Every one of those felt defensible on its own. Collectively they made the product impossible to describe in one sentence, which I've now come to believe is the leading indicator that a product is in trouble.
The interviews I've been running made it concrete. Founders kept describing two specific needs with real urgency. One was cold outreach at a scale they couldn't do manually, reaching fifty people a day with genuinely personalized messages, with almost no user effort beyond setup. Two was being present on the right surfaces at the right times without having to context-switch for each one. Everything else on my roadmap was wishlist material. Interesting, maybe useful, but nobody was leading with it.
The exercise this week is cutting. Every feature that doesn't serve the two founder jobs comes off the demo. Every capability that's technically impressive but isn't one of those two jobs gets shelved. That's uncomfortable because I'm proud of some of the things I'm cutting. The heartbeat layer, the active agent, the live transcription, all real engineering work. All of them stay in the code base. None of them lead.
Chinat pushed on why I'd been unwilling to cut. The honest answer is that breadth had been my psychological hedge. If the product can do ten things, one of them will hit, right? That's the story founders tell themselves when they haven't committed. The failure mode it produces is that you build a demo that does ten things badly and none of them spectacularly, and the audience walks away remembering none of it. Depth is what produces memory. Breadth is what produces polite nods.
The other frame that helped is thinking about the two jobs as distribution-shaped, not feature-shaped. Cold outreach at scale is not a feature, it's a wedge into founder distribution. If a founder starts using Pulse for outreach, they're using Pulse every day. That's the kind of habit that makes everything else we build cheaper to distribute, because we're already in the workflow. The second job, presence, compounds the same way. Once Pulse is responding on your behalf across the surfaces you live on, the surface area for future capabilities is already open.
The research side of my life is pushing the same discipline. The network-of-agents paper is narrowing to one specific claim rather than a framework. Agent networks become intelligent through structured interaction mechanisms, not through scale alone. The paper has to defend that claim and nothing else. Everything else is support structure. Same shape as Pulse's product cut. One main claim, everything else subordinate.
The part I'm sitting with is that the two founder jobs look small. Cold outreach is a thing other products do. Presence is a thing other products do. What's the moat? The honest answer is that the moat is depth, not category novelty. Our job is to do one or both of these so well that a founder comparing us to the generic alternative sees a clear gap. That's a different kind of moat than a technical novelty moat, and it's the kind I've been reluctant to lean on because it feels less academically interesting. But depth is what produces retention, and retention is what produces a business.
Chinat mirrored the same move from the MentorMates side. He's running a flagship hackathon in Hong Kong in July and he's treating that event as the forcing function for the whole product, not as a marketing moment. Same shape: one big test that forces everything to get sharp. In his case, an event with a thousand participants. In my case, a demo that does one or two things end-to-end with zero slop.
The last thing worth naming. I've been measuring progress by feature count, which is a bad metric for early products. The right metric is clarity. Can I describe Pulse in one sentence that a founder would react to? Can I demo it in ninety seconds in a way that tells a whole story? If not, I haven't cut enough.
The check I'm making on myself: by the next mastermind, the demo tells one story, end-to-end, and the paper has a one-page abstract that holds together. If the demo still has five use cases in it, I haven't done the work.