This morning, DiVine, the Jack Dorsey-backed reboot of Vine, became publicly available on the App Store, Google Play, and Zapstore. Most of the headlines are about nostalgia: 500,000 restored six-second videos, OG creators like Lele Pons coming back, the early-2010s internet getting another shot.
Nostalgia is the surface. The more interesting story is what diVine refuses to be.
The team, led by Evan Henshaw-Plath (“Rabble”), built the platform to filter out AI-generated content. New videos must either be recorded directly in the app or come with C2PA provenance, an open standard that establishes how a piece of media was created and edited. The architecture runs on Nostr, with experiments underway on AT Protocol and (eventually) ActivityPub. There is no revenue model. The company is a public benefit corporation.
In other words: an open-protocol, human-only video network, shipped at a moment when the rest of the feed is filling up with synthetic content.
We helped build it.
A hybrid model
Very Good Ventures came in shortly after the prototype phase. The initial codebase had been vibe-coded: moving fast, light on architectural patterns, exactly the kind of foundation that gets you to a working demo and then becomes hard to scale. Our job was to stabilize it, harden it, and prepare it for real users without killing the velocity that made it work in the first place.
Ninety days later, the team had addressed 500+ issues across the application, established consistent engineering patterns, and shipped to early users on Zapstore. Today’s public launch builds on that foundation.
What made it possible was a workflow built around Claude. Claude Code accelerated PR generation and review. MCP integrations automated ticket creation and backlog refinement. Project rules enforced code quality, Nostr protocol compliance, and brand consistency. A small cross-functional group of engineers, testers, designers, a PM, and a program manager operated with the leverage of a much larger one.
This is where the philosophy gets interesting.
The line that isn’t a contradiction
DiVine excludes AI-generated content from the user experience. We used AI extensively to build the user experience. Read quickly, that sounds like a contradiction. It isn’t.
The line diVine draws is between the tools and the output. We drew the same line. AI is allowed to make builders faster and the platform better. AI is not allowed to flood the feed with content that bypasses human creativity entirely. The first expands what people can ship. The second hollows out what’s worth shipping.
The way we put it in the case study:
The creativity is human. Claude just makes it ship faster.
That isn’t a slogan. It’s an architectural decision. It shows up in the product (C2PA verification, in-app capture). It shows up in the process (AI in the dev loop, not in the content pipeline). It shows up in the business model (PBC, no ad-driven optimization for attention).
Why it matters now
A lot of social platforms are sliding toward a world where users can’t tell what’s real, creators can’t compete with infinite synthetic supply, and the feed becomes a kind of statistical mush. DiVine is a small bet in the other direction.
It is also a useful proof point for how to build anything in 2026. AI acceleration is a question of where you put the leverage. Put it in the build pipeline and your team gets faster. Put it in the content pipeline and your product gets cheaper to fake.
DiVine put it in the right place. We were proud to help.
