Tracing the ghost in the machine.
Six months ago, a pre-seed round of $2 million led by General Partnership. Today, Sequoia doubles down with $10 million. The startup is Preview, an AI video production platform that claims to solve the mess of integrating generative AI into professional filmmaking. But the numbers alone don't tell the story. What caught my attention is a single line buried in the funding announcement: "Each frame records who generated it, what model was used, and the parameters applied."
That is not a feature. That is a ledger. And in a bear market where trust is the only scarce asset, that ledger might be more valuable than the video itself.
Context: The Quiet Ruin of Creative Provenance
We have been here before. In 2021, NFT marketplaces promised immutable provenance for digital art. The hype cycle delivered floor prices, but the actual metadata—who minted, what contract, which parameters—quickly became noise. Artists lost royalties, collectors bought fakes, and the promise of on-chain attribution decayed into gas wars. The problem was not the technology. The problem was that the market valued speculation over verification.
Now, AI video generation faces a similar crisis. Tools like Runway, Pika, and Sora produce stunning clips, but the production pipeline is a black box. A director cannot verify which model rendered a specific shot. A studio cannot audit whether a scene used copyrighted training data. A client cannot trust that the final cut was generated ethically. The industry is drowning in output, but starving for accountability.
Preview's pitch is to become the "central control panel"—a single workspace for scripts, storyboards, shot lists, AI generation, review, and feedback. Teams can use different models simultaneously, manage characters and scenes in a unified interface. That is useful. But the deeper insight is that Preview is building an audit trail for every pixel. The code remembers what the market forgets.
Core: The Narrative Mechanism of Provenance
Let me read the silence between the blocks. Sequoia's partner compared Preview to a "video version of Cursor." For those outside the developer loop, Cursor is an AI code editor that gained traction because it records every edit, every prompt, every model invocation. Developers use it not just for productivity, but for reproducibility. When a bug surfaces, they trace back through the chain of generations. Trust emerges from traceability.
Preview applies the same logic to video. Each frame carries a watermark of creation: who generated it, what model, what parameters. That is metadata. But metadata is only as powerful as the infrastructure that secures it. Right now, Preview stores that data in a centralized workspace. That is fine for a seed-stage product. But the natural evolution—the one that aligns with the narrative of Web3—is to anchor that metadata on a public, immutable ledger.
Imagine a blockchain where each frame is hashed, timestamped, and linked to a creator identity. A Hollywood studio could prove that a specific scene was generated using a licensed model. An advertiser could verify that a commercial complied with brand guidelines. An independent filmmaker could tokenize their workflow as a composable asset. The quiet ruin when the algorithm broke—that moment of lost trust—would be replaced by a verifiable chain of custody.
I have seen this pattern before. In 2022, after the Terra collapse, I spent months in Patagonia analyzing why algorithmic stablecoins failed. The root cause was not the code. It was the absence of a transparent audit trail for the collateral. The algorithm broke because the trust was abstract. Preview is not a stablecoin, but the same principle applies: if you cannot trace the generation, you cannot trust the output.
Quantitative Sentiment Forecaster: The data here is thin but telling. Preview already has over 100 studios using the platform, including agencies producing ads for Fortune 500 companies and Hollywood film production teams. Another 3,000 studios are on the waiting list. Assuming a conservative conversion rate of 30%, that is 1,000 active studios within the next year. Each studio generates dozens of frames per project. That is a massive dataset of provenance metadata. If Preview becomes the standard for professional AI video, it will hold the largest repository of generation metadata in the world. That is a data moat. But it is also a responsibility.
Contrarian: The VC-Manufactured Narrative Trap
Let me offer a counter-intuitive angle. The "video version of Cursor" narrative is seductive, but it is also a trap. Cursor succeeded because developers are already comfortable with version control systems like Git. The creative industry is not. Directors, editors, and producers do not think in terms of commits and branches. They think in terms of cuts and takes. The institutional narrative translator in me warns that Preview risks over-engineering a solution for a problem that most studios do not yet feel.
Sequoia's thesis is that AI video lacks a unified workflow. But the real bottleneck is not the tooling. It is the trust deficit between the creative team and the AI. Studios are hesitant to adopt generative AI because they cannot guarantee that the output will be consistent, legal, or safe. Preview's provenance tracking addresses that, but only if the studios adopt it. And adoption requires a behavioral shift that most creatives are not ready for.
Furthermore, the "omnichain app" narrative—that every product needs to be on every chain—is VC-manufactured. Preview does not need to be a blockchain company. It needs to be a great video production tool. The provenance layer can be additive, but it should not be the core value proposition. We traded chaos for consensus, and lost ourselves. If Preview becomes a blockchain-first platform, it will alienate the very studios it aims to serve. The smart contract doesn't care about the director's deadline.
Takeaway: The Next Narrative Cycle
So where does this leave us? Preview's $12 million raise is a signal, but not of a new crypto vertical. It is a signal that the market is beginning to value provenance over hype. The next narrative cycle will not be about AI video generation itself. It will be about the infrastructure that makes AI-generated content trustworthy. When the herd wakes, the signal has already faded.
For the studios waiting in line, the question is not whether Preview's platform works. It is whether they will be willing to pay for a trust layer that they currently do not have. For the investors, the question is whether Sequoia's bet will create a new asset class—provenance tokens, maybe—or just a better workflow tool.
I am watching the metadata. That is where the ghost lives. Finding community in the silence of the ape's gaze—the ape here is the AI, generating without understanding. The community is the network of creators, verifiers, and auditors who will build the trust layer. Preview is the first step. The next step is on-chain. And when that happens, the quiet ruin will become a quiet revolution.
