Welcome back! OpenAI just axed its own video platform six months after launch, and it's taking a billion-dollar Disney partnership down with it. Meanwhile Apple is making its biggest play for AI relevance yet, a robotics billionaire is building a new kind of hardware company and Meta is doing what everyone expected it to do three years ago.
In today’s Generative AI Newsletter:
OpenAI: Was Sora always a side quest?
Apple: Can a standalone Siri app save its AI reputation?
Hark: What does the iPhone Air designer want to build next?
Meta: Is the metaverse officially dead?
Latest Developments
OpenAI Kills Sora to Make Way for "Spud"
Six months after hitting No. 1 on the App Store, Sora is done. OpenAI confirmed it's winding down the video generation app, the API and all related video products to free up compute for its next major model, codenamed Spud.
What happened: Employees reportedly called Sora a "drag" on resources. The app, API and Sora.com are all being shut down, with timelines and content preservation details coming soon.
The Disney fallout: The $1B investment and 200-character licensing deal announced in December is dead. Disney learned about the wind-down 30 minutes after a working meeting with OpenAI.
Where the compute goes: Spud has completed initial development and Altman told staff it "can really accelerate the economy." The Sora research team will pivot to world simulation for robotics.
Leadership reshuffle: Altman is stepping back from safety oversight entirely. Mark Chen takes over safety, Greg Brockman handles security, and Altman focuses on fundraising and data centres.
This is OpenAI in full pre-IPO mode. Everything that isn't ChatGPT, Codex or the planned super-app is getting cut. The fact that Sora burned through GPU resources while Anthropic was shipping productivity tools tells you where the competition is heading.
Special highlight from our network
For the last ~3 years, the AI industry has been racing to build bigger, smarter models like ChatGPT, Claude, Gemini and so on. The assumption was simple: whoever built the most powerful model wins.
But that race is actually already slowing down.
Now the industry is beginning to look beyond the model race and ask: how do companies turn AI into something that actually works and brings value?
In this LinkedIn Live, Ori Goshen (Co-Founder & Co-CEO of AI21 Labs) joins Steve Nouri to unpack what the Post-Model Era means for enterprise AI, from moving beyond the model race to building systems that are reliable, practical and ready for production.
Wondering where AI is heading next? This conversation will help put the direction into perspective.
Join LinkedIn Live to find the answers.
Apple Rebuilds Siri as a Standalone AI App
According to Bloomberg's Mark Gurman, Apple is testing a dedicated Siri app and a new "Ask Siri" chatbot experience, both set to debut at WWDC on June 8 with iOS 27.
What's new: Siri gets its own app for the first time with a redesigned interface supporting both text and voice input.
Context awareness: The assistant will reportedly read across iMessages, emails and notes to build context and execute actions inside third-party apps.
The rebrand: Apple is positioning the experience as "Ask Siri", a chatbot-style interaction replacing the rigid voice commands of previous versions.
The engine: The whole thing runs on Google's Gemini model through Apple's Private Cloud Compute infrastructure.
After Apple Intelligence landed with a thud last year, this is probably their last real shot at keeping users from defaulting to ChatGPT or Claude for everything. June 8 just became the most important Apple keynote in years. Whether the demos match reality this time is another question entirely.
Figure AI Founder Launches $100M AI Device Lab
Brett Adcock, the CEO of $39B robotics company Figure AI, has officially launched Hark. The new AI lab is building personalised intelligence paired with dedicated hardware, and Adcock has put $100M of his own money in.
The team: 45 people pulled from Apple, Google, Meta and Tesla. Hardware design is led by Abidur Chowdhury, who previously led design on the iPhone Air.
The vision: A "family of devices for yourself and the home" with persistent memory, multimodal input and systems that "begin to think like you and sometimes ahead of you."
The compute: Thousands of NVIDIA B200 GPUs arriving in April, with the first AI models and software planned for this summer.
The approach: Full vertical integration. Models, hardware and interfaces built together from scratch, not bolted onto existing platforms.
The AI device space has been brutal. Humane's AI Pin flopped, Rabbit's R1 turned out to be an Android app in a box and OpenAI's Jony Ive collaboration has been pushed to 2027. Adcock's pedigree gives Hark more credibility than most, but he's still entering a category where nobody has cracked the code yet.
Meta Finally Admits the Metaverse Is Over
Meta shut down the VR version of Horizon Worlds last week, capping off roughly $80B in cumulative Reality Labs losses since 2020. The company is now pouring that money into AI agents instead.
The shutdown: Horizon Worlds is off the Quest store as of March 31 and fully removed from VR on June 15. Only a mobile app survives.
The agent push: Meta acqui-hired AI agent startup Dreamer this week, its third agent-focused deal in four months following Manus and Moltbook.
Internal adoption: Meta employees are reportedly flocking to MyClaw, a personal agent built on OpenClaw. Some are giving their agents access to chat logs and work files to run 24/7.
The spend: $135B in capital expenditure planned for 2026, nearly double what they spent in 2025, with the vast majority going to AI infrastructure.
Zuckerberg renamed the entire company for a vision he's now abandoning less than five years later. The metaverse was never wrong as a concept. It was wrong on timing, wrong on hardware and wrong on the assumption that people would strap a headset to their face for social interaction. The pivot to AI agents is the right call. It's just a (very) expensive admission.
Tool of the Day: Moda
Moda is an AI design agent that learns your brand and creates fully editable presentations, social posts and marketing assets on a collaborative canvas. It just launched publicly after raising $7.5M from General Catalyst and Pear VC. You describe what you need, three agents handle research and create a draft, and you step in to fine-tune. Everything exports to Google Slides and PowerPoint. Free to start with 1,000 AI credits.
Try it yourself at moda.app
