Six months ago, Sora was the most exciting app on the internet.
OpenAI had just signed a deal with Disney, Marvel, Pixar, and Star Wars, all of which are available inside an AI video app. Sam Altman called it a "cinematic sea change." Downloads shot to the top of the App Store. Sora was going to be the AI-native TikTok.
On Tuesday, OpenAI shut it down.
No warning or explanation. Just a brief post: "We're saying goodbye to the Sora app." Disney found out the night before it was announced. Their planned $1 billion investment in OpenAI? Gone with it.
So what actually happened?
Sora had a deepfake problem from day one. Within weeks of launch, users figured out how to bypass the guardrails. Realistic videos of Martin Luther King Jr., Robin Williams, and Sam Altman himself were being generated and shared. Families of deceased public figures had to post on Instagram asking people to stop. It was a moderation nightmare with no end in sight.
Then the numbers came in. Downloads fell nearly 75% from their November peak. OpenAI was burning expensive GPU compute, its scarcest resource, to generate videos that almost nobody was watching twice.
Meanwhile, Anthropic was quietly capturing 73% of new enterprise AI spending. Coding and workflow tools. Boring, practical, high-margin work.
That's when OpenAI finally looked at Sora and asked the question it had avoided: what is this actually for?
The real answer is the IPO. OpenAI is heading toward a public listing and needs to show investors a clear, profitable business, not a content moderation liability burning cash to make deepfakes of celebrities. Enterprise clients don't want video generators. They want tools that improve output and reduce costs. Anthropic gave them that.
OpenAI is now pivoting hard to catch up.
Sora's underlying model isn't gone. It's being redirected toward robotics and "world simulation,” which means exactly what it sounds like: teaching AI to understand physical reality by generating it.
The hype era is over. And the business era has started.
Claude can now use your computer. And soon, your phone.
Text Claude a task from your phone. Come back later and it's done. Anthropic just launched this for Claude Pro and Max users. And buried in unreleased code, a feature called Orbit was spotted: Claude making calls, sending messages, and managing apps directly on your smartphone.
It’s fully autonomous, no desktop needed. The gap between "AI assistant" and "AI employee" just got a lot thinner.
Bank of America just put AI agents inside financial advisor chairs.
The bank deployed an AI advisory platform to 1,000 financial advisors this week. It handles client queries, prepares recommendations, and manages daily workflows in real time, alongside the human.
For context, BofA already says its AI virtual assistant “Erica” does work equivalent to 11,000 employees. All 18,000 of its software developers use AI coding tools.
Kleiner Perkins just raised $3.5 billion for AI only.
One of Silicon Valley's oldest and most respected VC firms closed its largest fund ever and made it explicit: every dollar goes to AI. When a 54-year-old firm that has backed Google, Amazon, and Genentech decides to go all-in on a single theme, it's a conviction bet.
The money flowing into AI isn't slowing down. It's concentrating.
FAST BREAK
Last week, Sony Music requested the removal of 135,000 AI-generated songs from streaming platforms. In one week. That's not a piracy problem but a content flood.
Spotify just launched "Artist Profile Protection."
Artists can now approve or reject any track before it appears under their name. Because the platform had no way to stop fake AI music from landing on real artists' pages and counting toward their stats.
The internet took 20 years to figure out how to fight fake news. AI gave the music industry the same problem in 18 months.