The App That Stopped Working
Sora's consumer app shuts down today. The tile is still on a lot of phones. Tap it and you get a migration screen pointing you at ChatGPT. Existing libraries are read-only. The API follows on September 24. That's the lifecycle.
A year ago a lot of designers were calling Sora a category. Today it's a deprecation notice. GPT-5.5 absorbed the territory inside ChatGPT, the same way GPT-5.5 absorbed image generation, code execution, and most of the other surfaces OpenAI used to ship as separate products. The "superapp" framing TechCrunch put on the GPT-5.5 launch is the operating thesis. One model, one surface, everything else gets folded in or sunset.
Two reasonable reactions to that. The first is outrage. The second is "of course." I'm in the second camp.
Software Gets Sunset
The thing some people forgot the last eighteen months is that AI products are software. Software gets sunset. That's not a betrayal. That's how the category has worked since before any of us had a portfolio.
Adobe killed Flash. Apple killed iWeb. Google killed Inbox, Reader, Stadia, and roughly fifty other products with a flag at the top of the page that said "this is the future of how you do this thing." Figma absorbed Adobe XD's user base by being better, and then Adobe walked it back. Every designer who has been working long enough has watched at least one tool they relied on get retired.
What's different with AI is the speed. Sora launched in early 2025. It got broad consumer access in late 2025. It's gone in April 2026. That's a sub-eighteen-month run for a product that, at one point, had Hollywood writing position papers about it.
The speed is the actual story. Not the death. The product life cycle compressed.
What I Actually Built On Sora
I'll be honest. Not a lot. I tested it, made a few clips, ran storyboards through it for fun, never built a billable workflow around it. That's not because I'm clever. It's because I was already building on Veo and Seedance for paid work, and Sora's library wasn't where my output was living.
But I have friends who did build on it. Solo creators with B-roll pipelines, small studios using it for pitch animatics, one indie ad shop who was rendering treatment videos for client meetings. They're spending today migrating prompts and storyboards into other tools. Veo 4 launched April 15. Seedance 2.0 went live inside HeyGen on April 7. The tools are there. The work moves.
The migration is what's interesting. Because what they're moving is not the model. They're moving the input.
The Input Is The Asset
This is the part that matters for any small business owner reading this who's been quietly worried that their AI tool of choice might disappear.
The thing that has value, and the thing that survives a sunset, is the input. The prompt. The storyboard. The brand spec. The reference image stack. The shot list. The voice guide. The character description. The thirty iterations you ran to get a hand to actually look like a hand.
That's portable. A storyboard for a Sora clip is a storyboard for a Veo clip is a storyboard for whatever ships in eight months. The model on the other end of it is the variable. Your specification is the constant.
I learned this on a different scale, building Ruck Authority. I have a video pipeline that turns articles into branded videos with a digital avatar and a voice clone. The pipeline has six stages. Three of them have been swapped for different vendors in the last year. The thing that didn't change is the script template, the brand spec, the avatar reference, and the way I want shots to cut. Those are mine. The vendors are interchangeable.
A lot of people reading the Sora announcement today are reading it as "the platform I bet on is dying." That's a reasonable feeling and also the wrong frame. The platform was always going to be a renter. The brand spec, the prompt library, the storyboard, those are the deed.
What To Do This Week
If you've been generating video through Sora for any kind of business reason, here's the practical version.
Export everything you can while the library is still readable. The libraries become read-only today, which means you can still see them. There's no public commitment that the read-only state lasts forever. Pull the references you want to keep.
Save the prompts as text. Not screenshots of the Sora interface. Actual prompts, in a doc, with notes about what worked. A prompt is markdown. Markdown survives.
Pick a new tool, but pick it for output, not for vibes. Veo 4 and Seedance 2.0 are both viable. Runway is still here. So is Pika. Generate the same shot in two of them and see which one renders the brand the way you want it. The model that ships your spec wins this week. It might not be the model that wins next quarter.
Write down your brand video spec somewhere outside any one tool. Brand colors, typography choices for any title cards, the kinds of cuts you favor, the shot lengths, the voice if you use one. Keep it as markdown. This is the file you migrate. Not the videos. The spec.
The Real Sunset Isn't The App
The Sora app is gone today. That's the headline.
The thing I'm watching is the next layer down. AI tools are starting to behave like the rest of the software stack, which means consolidation, deprecation, and migration are now part of the work. Same as it ever was, just on a faster clock.
The designers who survive this cleanly aren't the ones with the biggest tool stack or the earliest beta access. They're the ones who treat every tool as a renter and own the brand layer themselves. Specs. Prompts. Storyboards. The stuff that travels.
I built joel.design and Ruck Authority and a half dozen other things on the assumption that any individual tool in the pipeline could go away tomorrow. Today one of them did. The work continues because the work was never in the tool.
That's the lesson. Own the input. Rent the rest.