Taking a vibe-coded short-form video app to production-ready release.

Divine is a new take on ultra-short-form video — built around 6-second clips designed for the modern attention span. The initial prototype was rapidly “vibecoded” by Evan Henshaw-Plath (Rabble), drawing on a large public archive of early short-form internet video content to help bootstrap a decentralized, open-source social video platform.

From the outset, Divine has operated with a clear ethos: preserve human creativity in the product experience, while using advanced technology to accelerate the building of that experience.

Very Good Ventures (VGV) joined shortly after the prototype phase to help transform a fast-moving proof of concept into a stable, production-ready application. Over the course of 90 days, VGV worked alongside the founding team to introduce structure, strengthen the codebase, and prepare the app for real users — without losing the speed and creative momentum that defined its origin.

The initial codebase had been developed rapidly, without established architectural patterns or consistent state management. While this enabled speed, it created challenges for stability, scalability, and team coordination as the product matured.

At the same time, the product itself was still evolving. The founding team brought deep expertise in Nostr — the decentralized protocol powering the platform — along with a clear vision for what Divine could become. However, translating that vision into a production-grade Flutter application required additional engineering rigor, platform-specific expertise, and coordinated workflows.

Development, design, and product decisions were happening in parallel across a distributed team. New ideas and features were continuously being introduced even as stabilization efforts were underway. This created a dynamic environment where priorities shifted quickly and required constant alignment.

Adding to this complexity was a deliberate product constraint: while AI would be used extensively in development, generative AI video content would not be part of the user experience. Maintaining that boundary required clarity in both product and technical decision-making.

This was not a traditional “stabilize and ship” engagement. It required building structure in real time — while preserving the velocity, experimentation, and intentional product ethos that made the platform compelling in the first place.

Divine operates along a deliberate boundary: AI is foundational to how the product is built — but not to the content it serves.

The platform is designed to prioritize human-created expression. Generative AI video content is intentionally excluded from the user experience to preserve authenticity, creativity, and signal in a space increasingly saturated with synthetic media.

At the same time, AI — particularly Claude — is deeply embedded in the development process.

This is not a contradiction. It is a design choice.

Claude, itself a human-created system, is used by builders to expand what they are capable of creating — accelerating development, improving quality, and enabling a small team to operate with the leverage of a much larger one.

The division is clear:

• Humans create the content
• AI accelerates the creation of the tools

This reflects a broader ethos behind Divine and its parent ecosystem — the And Other Stuff (AOS) project: technology should expand human agency, not replace it.

Claude isn’t creating on Divine — it’s helping build Divine.