Jack Dorsey, the co-founder of Twitter who killed Vine in 2017, is now backing its return as diVine, a six-second video platform with a strict ban on AI-generated content, the Guardian Project’s AI detection tools built into the upload flow, and the founding proposition that human creativity is an entire product, not a feature.
The app is live. Launched to testers last November under Dorsey’s nonprofit organization and Other Stuff, and going public in May 2026, it currently operates with an archive of about 500,000 videos from the original Vine, as well as new submissions from human creators. The lead operator is Evan Henshaw-Plath, known online as Rabble, an early Twitter employee who has been building Vine’s content library into a permanent home for several years. Henshaw-Plath’s argument is that the original Vine represented a particular kind of Internet creativity that was constrained, fast, looping, and human, and that both the age of algorithmic feeds and the age of generative AI are eroding in various ways. diVine is an attempt to rebuild that format under a set of rules that will prevent erosion from happening again.
Enforcement mechanisms are important because while any platform can create a policy that prohibits AI content, few can reliably implement it. diVine requires new uploads to be verified as human authored, either directly within the app or through a third-party verification tool. The Guardian Project, a privacy and security software organization, provides an AI detection layer that flags potentially synthetic content before it’s posted. This is a more specific infrastructure investment than community guidelines, but it’s not foolproof. Although AI video detection is becoming more accurate, false positives and negatives still occur, frustrating both legitimate human authors and determined malicious actors. The honest version of DiVine’s content credibility pitch is that it significantly increases the cost and friction of posting AI-generated content, and it doesn’t completely rule out that possibility. For a platform of this size, the difference is manageable. At the scale DiVine needs to reach to meaningfully compete with TikTok and Reels, that becomes a core operational and technical challenge for the business.
Competitive positioning for AI-generated content is a real market signal, not pure nostalgia marketing. Pew et al.’s research consistently shows that users of all ages have a skeptical attitude toward AI-generated media, especially in entertainment and social contexts where authenticity and relevancy are key value factors. Short-form video is the format in particular where skepticism is most acute. The appeal of a six-second loop has always been the feeling of a real person doing something interesting, interesting, or strange in front of a camera and sharing it without the production apparatus of traditional media. Sora, Runway, and Kling have made reliable 6-second AI-generated videos cheap enough that platforms without enforcement mechanisms will be flooded with synthetic content within months of achieving meaningful scale. diVine’s founders understand that the authenticity of the human-only label is what sells them, and the moment the label is convincingly challenged, the core value proposition collapses.
The Dorsey angle is more complicated than reports suggest. Dorsey left Twitter in 2021 and has since established himself as a critic of the centralized platform model he helped build, investing in decentralized social infrastructure through Bluesky and related projects. Supporting DiVine through a nonprofit rather than a for-profit organization is consistent with its positioning, but also raises questions about its business model. diVine’s guidelines mention integration with Dorsey’s block ecosystem for potential monetization, which refers to creator payments and tipping infrastructure rather than advertising as the revenue mechanism. While this approach is philosophically consistent, it has not yet been proven to generate enough revenue to simultaneously sustain content management teams, discovery technology partnerships, server infrastructure, and competitive creator acquisition budgets. The original Vine didn’t solve creator monetization problems, even with 100 million monthly active users and Twitter’s balance sheet. diVine is starting with a fraction of its audience and has no advertising engine.
The broader pattern that DiVine fits into is consumer startups valuing authenticity as a response to AI content saturation. The company isn’t alone in making this bet. BeReal built its brief cultural moment on a similar premise: that Instagram’s selective inauthenticity is a problem worth solving through product constraints. Rocket Widget, Poparazzi, and a wave of ephemeral friend-first social apps have all experimented with format and audience constraints as alternatives to the algorithmic feeds that dominate existing apps. None of them scale up to durable platforms, and a common failure mode is that the constraints that create initial differentiation become caps for growth once the initial group of enthusiasts is exhausted and mainstream creators demand the flexibility and monetization tools they already have on TikTok and Reels. diVine’s 6 second limit and AI ban are both constraints in this sense. They create identity and early loyalty. It also limits the world of creators and the surface area of content that drives discovery and retention.
The question for founders watching this launch isn’t whether diVine will beat TikTok. That won’t happen. The question is whether anti-AI positioning is a bankable thesis for a new category of consumer platforms, and whether that category has enough of an audience to support a standalone business. Evidence from search trends, platform research, and early creator responses to DiVine’s launch suggests that the audience for human-only content is real and engaged. Whether it’s large enough, loyal enough, and willing to pay enough to sustain the platform through the content management costs required to honestly enforce a humans-only policy will be the calculation that will determine whether diVine becomes a durable product or a well-intentioned experiment to demonstrate demand without knowing the economics.
Also read: IBM’s MAMMAL is a quiet demonstration that biomedical AI is moving beyond single-purpose models • Panthalassa wants to build AI data centers at sea, but power shortages make it sound less insane than it needs to be • AI models are giving biosecurity experts guidance on operationally useful biological weapons, and denial systems aren’t stopping them
