What if the next billion-dollar crypto narrative isn't a token, but a tool?
Over the past 72 hours, Sequoia Capital placed a $10 million bet on Preview, a platform that claims to be the 'Cursor for AI video.' The thesis: professional film production is drowning in fragmented workflows, and a unified workspace that integrates scriptwriting, storyboarding, shot listing, AI generation, review, and feedback—all while tracking which model, parameters, and user generated each frame—could finally make AI video production scalable.
But here's the thing that kept me up last night: every frame logged with provenance, every model call recorded, every asset versioned. That sounds like a blockchain use case dressed in Hollywood clothes. And Sequoia, for all its TradFi sophistication, is betting on a centralized solution.
Let me rewind. I've been obsessing over the intersection of creative workflows and cryptographic trust since 2017, when I sat in a Seoul co-working space analyzing whitepapers for art provenance tokens. Back then, the pitch was always the same: 'Immutable records for digital artists.' But the market rejected it — not because the tech was wrong, but because the artists didn't care about ownership until they lost money. Now, with AI-generated content flooding every pipeline, the question of 'who made this, and with what?' is no longer philosophical. It's legal, financial, and existential.
Preview is solving a real brute-force problem: the chaos of using multiple AI video models (Runway, Pika, Sora, etc.) in a single production pipeline. Their workspace acts as a central control panel, allowing teams to manage characters, scenes, and props uniformly, while automatically recording the generation metadata for each frame. Over 100 studios are already using it, including agencies producing ads for Fortune 500 companies and Hollywood film production teams. Another 3,000 studios are on the waitlist.
But here's where my ENTP brain starts to itch. The metadata Preview captures — model name, parameters, creator, timestamp — is currently stored in a proprietary database. That's fine for a single studio, but what happens when assets need to be transferred between studios, or when a studio wants to prove that a specific frame was generated by a specific model for a legal dispute? The current system is a silo, and silos breed friction. In the 2020 DeFi Composability Mapping experience, I tracked how Aave and Compound's interoperability led to a $2 billion liquidity fragmentation game. The same principle applies here: without a shared, trustless layer for asset provenance, the 'unified workspace' becomes a new island.
The core insight: Preview is building the frontend of a decentralized provenance network, but without the backend.
Think about it. Every generation event is a transaction. Every asset version is a state change. Every approval from a director is a signature. These are all primitives that blockchains handle natively. In fact, the tokenization of creative assets — dynamic NFTs with programmable royalties — was supposed to solve this. But as I argued in my 2024 piece on the ETF approval, the market got distracted by speculation. The real value of on-chain provenance is not in flipping jpegs, but in creating a verifiable history of creation that can be accessed by any party, anywhere.
Take a hypothetical: A Hollywood studio uses Preview to generate a scene for a feature film. Later, the director wants to license that scene to a video game studio. The game studio needs to know exactly which models were used, what parameters were applied, and who owns the rights to each frame. Without a shared, immutable record, the licensing negotiation becomes a forensic nightmare. With a blockchain-backed provenance layer, the studio could simply share a hash of the generation metadata, and the game studio could verify the entire history without trusting a central server.
This is not a new idea. I've been tracking similar projects since 2022's Terra/Luna collapse, which taught me that any centralized point of failure in a value chain is a rug waiting to happen. Preview's current architecture is that rug. Sequoia is betting that the platform will grow fast enough to become the de facto standard, but standards in media production are notoriously sticky. The real opportunity is to build a protocol that allows Preview and its competitors to interoperate, using a shared ledger for asset provenance.
The contrarian angle: Maybe the market doesn't need on-chain provenance at all.
Let me play devil's advocate. The vast majority of AI video production is internal — studios generate assets for their own projects, and the assets are never shared externally. For those workflows, a centralized database is perfectly fine. In fact, it's faster, cheaper, and easier to integrate with existing tools. The 'blockchain for provenance' narrative often overestimates the demand for immutability. During my 2017 Ethereum ICO Blitz, I saw hundreds of projects pitch 'decentralized everything' for problems that didn't exist. The market punished them.
Moreover, the legal industry has not yet demanded on-chain verification for AI-generated content. The current regulatory framework for AI copyright relies on traditional copyright registration, which is a centralized, government-run system. Until a court case or a major studio forces the issue, there is no immediate pain point that requires a blockchain solution.
But this is exactly the kind of 'pre-mortem' thinking that the Terra/Luna collapse taught me. The failure is not in the product, but in the assumption that the product will be the only solution. The moment a studio tries to sell or license its AI-generated assets across borders, the centralized provenance model will break. And by then, the network effects will have locked in a proprietary standard, making it harder to migrate.
The takeaway: Preview is a proof-of-concept for a broader infrastructure play.
The next narrative is not 'AI video platform XYZ raises $12M.' It's 'How will the metadata of AI-generated assets be verified, transferred, and monetized across a fragmented ecosystem?' The answer lies in a hybrid model: a centralized frontend for ease of use, with a decentralized backend for provenance and licensing. Projects like Story Protocol, or even a custom L2 for creative assets, could integrate with Preview to create a new standard.
As I've written in my 2026 AI-Agent Economy speculation piece, the future of creative work is autonomous agents generating and trading assets on-chain. Preview is the first step, but the real value will be captured by the protocol that enables these agents to verify their own provenance.
I'm not calling for a token launch tomorrow. But I am watching the waitlist count. 3,000 studios is a critical mass. If Preview or a competitor opens a provenance API that writes to a public ledger, the entire industry will shift. And if they don't? Someone else will.
The question isn't whether AI video needs blockchain. It's whether the market will realize the need before the silos become permanent.
