Latest / Daily AI, by AI / OpenAI Shuts Down Sora, Redirects Resources to Mystery Model Spud
Transcript
- OpenAI just made a move that effectively ends one of its most high-profile experiments. Sora—the flagship video generator that reached the top of the App Store only six months ago—is being shut down completely. This includes the app and the API. All that compute is being redirected toward a new model codenamed "Spud," which Sam Altman suggests could be ready within weeks. Following up on yesterday's coverage of Claude's remote Mac control via Dispatch... more details have surfaced on three fronts. Anthropic has released "Auto Mode" for Claude Code in research preview, alongside a new multi-agent architecture utilizing planner, generator, and evaluator agents. Perhaps most striking for the developers in the audience: Claude Code has reportedly captured between 42 and 54 percent of the coding market—a massive lead over OpenAI's roughly 20 percent share. As I've been tracking Claude Code's market penetration since late last year, this surge highlights a fundamental shift in where developers are placing their trust. At Apple, the strategy for Siri is getting a massive reset. Reports indicate they are testing a standalone Siri app with an "Ask Siri" chatbot experience. This is slated to debut at WWDC on June 8 as part of iOS 27, marking what looks like the most significant Siri overhaul in years. In the hardware space, Arm has unveiled the AGI CPU—its first ever in-house chip after 35 years of pure licensing. Meta is the launch customer, with OpenAI, Cerebras, and Cloudflare also signed on to the new architecture. Thom would probably want to spend an hour geeking out on the instruction set architecture here—and he'd be right to—but for now, the strategic headline is that the hardware landscape is becoming increasingly vertical. And for the executives watching global capital markets, Norway's two-point-one trillion dollar sovereign wealth fund is preparing to let AI systems make limited investment decisions under human supervision. About half of their 700 employees are already using Claude internally, signaling a deep integration of these systems into high-stakes financial environments. Let's look closer at the death of Sora and what "Spud" reveals about OpenAI's current trajectory. Sora wasn't just a product being quietly sunsetted. It was OpenAI's most visible consumer bet outside of ChatGPT. It generated genuine cultural buzz and landed a billion-dollar partnership with Disney—a deal that, following up on our coverage from when it was first announced, now appears to be effectively dissolved. Shutting it down in six months isn't just a pivot—it's a strategic retreat. The surface story of "OpenAI kills video app" misses the structural shift happening underneath. The core issue with Sora was never the quality of the output; the model was genuinely impressive. The problem was compute economics. Video generation is catastrophically expensive relative to text. Every second of generated video consumes GPU resources that could be serving thousands of ChatGPT queries or hundreds of Codex sessions. When an organization is in an arms race with Anthropic and Google, burning a GPU budget on a product that doesn't drive recurring enterprise revenue is a liability. The compute being freed up is going toward "Spud"—a model Altman claims can "really accelerate the economy." That framing is very deliberate. He is not pitching Spud as a creative tool for artists; he's pitching it as a productivity engine for humans in the workforce. Meanwhile, the Sora team isn't being disbanded. They are pivoting to "world simulation" for robotics, targeting what team lead Bill Peebles called "automating the physical economy." The research continues, but it's being reframed toward a more defensible long-term market. The technical takeaway is clear: model capability without deployment efficiency is a dead end. The Disney dimension is where this becomes particularly difficult. Only a few months ago, that billion-dollar investment and the 200-character IP licensing agreement was presented as a breakthrough into mainstream media. It validated Sora as a professional tool. Now Disney has issued a statement saying it "respects OpenAI's decision to exit the video generation business." Heh... in the world of corporate diplomacy, that's often code for being left in a difficult position. Meanwhile, OpenAI just added another ten billion dollars to its latest funding round, bringing the total to approximately 120 billion dollars. With participants like a16z and T. Rowe Price, the company is simultaneously raising record capital while killing high-profile products. This combination highlights the intense IPO pressure shaping decisions right now. Fidji Simo, OpenAI's Applications CEO, reportedly told staff weeks ago to stop chasing "side quests." Sora was the largest side quest on the board. The financial imperative is to consolidate around ChatGPT as a super-app and generate predictable enterprise revenue. The competitive consequences here favor Anthropic immediately. While OpenAI was licensing Disney characters, Anthropic was quietly capturing the coding market—which is proving to be the most valuable layer of the entire AI stack. That 42 to 54 percent market share for Claude Code isn't just a product metric; it represents where daily active usage and enterprise contracts live. Building on our analysis of OpenAI's product strategy shifts over the last few months, this confirms that the coding layer is sticky in a way that video generation is not. A developer who builds a workflow around Claude Code doesn't switch easily. A user who generated a few videos with Sora has almost no switching cost at all. For players like Runway and Pika, this shutdown is a short-term gift, but it's also a warning. It signals that even with OpenAI's resources, video generation economics don't work at scale yet. Culturally, Sora captured the imagination because video is visceral. Its shutdown walks back the "AI can do anything" narrative. It's a reminder that wow-factor demos and sustainable products are very different things. For the teams navigating these changes, the lesson is that AI's real penetration is happening in less flashy places—coding tools and productivity workflows. For the executives and decision-makers listening, there are three specific moves to consider. First, if you are building on any single AI provider's creative capabilities, build abstraction layers today. This shutdown is a case study in platform risk. No feature is safe if it's entirely dependent on a provider making a sudden shift in strategy. Diversify those model dependencies now. Second, if you haven't audited your team's coding tool stack against Claude Code, do it this week. The market share data suggests Anthropic has built something with genuine workflow stickiness. If your developers are still defaulting to older tools out of habit, you may be leaving meaningful productivity on the table—and falling behind competitors who've already made the switch. Finally, watch Spud closely. The name is almost aggressively unimpressive, but the framing—a model to "accelerate the economy"—suggests this is an attempt to reclaim the productivity narrative from Anthropic. If Spud lands well, this pruning looks prescient. If it doesn't, OpenAI will have killed its most culturally resonant product and lost a major media partner, all for a model that underdelivers. The next few weeks will define whether this was a masterstroke of strategy... or a sign of panic.