Sora Wasn't a Product. It Was a Demo With a Download Button.
Sora hit a million downloads in five days. Six months later, it disappeared.
Not because nobody wanted it. Because nobody came back.
The spectacle problem
The launch was massive. Millions of downloads. Viral clips flooding every platform. The technology was genuinely impressive. Type a prompt, get a cinematic video. It felt like the future arriving ahead of schedule.
Then usage fell off a cliff.
This is what happens when you design for the first session, not the tenth. Sora was built to impress. And it did. You typed a prompt, got a stunning video, shared it on social media, and then... what?
No loop. No workflow. No reason to open it tomorrow.
It was a magic trick. And magic tricks only work once. The second time you see the rabbit come out of the hat, you're already thinking about lunch.
Spectacle vs. utility
There's a fundamental tension in product design between making something impressive and making something useful. The best products manage both, but if you have to choose, utility wins every time.
Spectacle drives downloads. Only utility drives retention. Figma is learning this the hard way with AI features that impress but don't stick.
Instagram wasn't impressive because it could apply a filter. It was useful because it gave you a reason to open it every day. Your friends were posting. Slack wasn't impressive because of its interface. It was useful because your team's conversations lived there.
Sora had the spectacle nailed. The technology was state of the art. The outputs were cinematic. The demos were viral. But there was no utility underneath.
Nobody wakes up on a Tuesday and thinks, "I need to generate a cinematic video today." Content creators, the most obvious user base, need specific shots, consistent characters, matching styles across a project. Sora gave you a one-shot magic trick, not a production tool.
The economics of empty retention
The costs kept climbing. Video generation is computationally expensive. Every prompt burns significant GPU time. When users aren't coming back regularly, the cost-per-active-user skyrockets.
Deepfakes flooded the platform. Copyright headaches piled up. Moderation costs added another layer of expense on top of already-high compute costs.
When a product can't justify its own compute costs through daily use, it's already dead. The Yellow Pages is the analogue version of this: a product that keeps shipping despite having lost its reason to exist. The revenue model depends on habitual users, not one-time tourists.
The lesson for every AI product
OpenAI didn't kill Sora because the technology failed. They killed it because no one designed a reason to stay.
This is the challenge facing every AI product right now, and $2.5 trillion in AI spending isn't addressing it. The technology is genuinely capable. The demos are legitimately impressive. But demos don't build businesses. Products do.
The difference between a demo and a product is the answer to one question: what does the user do on day ten?
If you can't answer that question before launch, you don't have a product. You have a demo with a download button. And the best demo in the world is still just a demo if nobody builds a product around it.
---
