OpenAI killed Sora. The question worth asking is why they built it.
Six months after Sora topped the App Store, OpenAI shut it down. Killed the app, killed the API, walked away from a $1B Disney deal in the process. The official reason is compute priorities. The real lesson might be about the cost of building things for the wrong reasons.
The Main Story: OpenAI Prunes for the IPO
What happened: OpenAI shut down Sora, its AI video generator, citing the compute drain during a period of intense competition. The Disney partnership, which had included a $1B investment and a 200-character IP licensing agreement, is now on ice. Sam Altman handed off safety oversight to focus on fundraising and data center buildout. OpenAI is completing a new model codenamed “Spud” and raising another $10B, bringing its latest round to roughly $120B total.
Why it matters: This is not seven separate stories. It is one decision played out across several announcements: OpenAI is stripping everything that is not the core ChatGPT super-app ahead of its IPO. Video generation was expensive, copyright-entangled, and apparently not something users chose over competitors. Shopping flopped. Safety oversight got delegated. What remains is a focused bet on ChatGPT plus Codex plus whatever Spud turns out to be.
The TWO angle: The detail worth sitting with is the Disney deal. Three months ago it was OpenAI’s most visible mainstream partnership. Now it ends with a press statement where Disney says it “respects OpenAI’s decision.” That word, respects, is doing a lot of work. OpenAI spent real resources building Sora into something mainstream enough to land that deal, then determined it was a distraction. Ecclesiastes says it plainly: “There is a time to plant and a time to uproot” (Ecclesiastes 3:2). The lesson is not that video AI is dead. It is that building something impressive and building something worth keeping are different problems entirely. The wise builder counts the cost before laying the foundation (Luke 14:28). Anthropic, meanwhile, has been adding Claude Code capability and shipping Cowork while their competitor was managing a Disney exit. Focus compounds.
The Rest of Today
Claude can now run tasks on your Mac while you’re away. Anthropic’s Dispatch feature, included in Pro and Max plans, lets you send Claude a task from your phone and have it execute on your desktop, with full file access, running in the background. This is not a demo feature. For a solo operator, it means handing off a repeatable task before you leave for the day and coming back to work that is done.
Claude Code added an auto mode for hands-free permission handling. If you have been using Claude Code and pausing constantly to approve individual steps, the new auto mode removes that friction. The tool proceeds through multi-step tasks without asking permission at each stage. More on this in the tool section below.
Figma opened its canvas to AI agents. Claude Code and other tools can now create and edit designs directly on the Figma canvas using a team’s existing components and brand standards. For marketers and non-technical operators building anything that requires designed assets, this collapses one more wall between “I described what I want” and “it exists.”
Arm built its first chip in 35 years. The company that licenses the designs inside nearly every smartphone finally built its own processor, a 136-core unit aimed at AI inference. Meta is the launch customer, with OpenAI, Cerebras, and Cloudflare also signed on. This matters for builders because the cost of running AI inference is directly tied to how efficient the underlying hardware is. More competition in chips means lower costs over time.
Apple is apparently serious about Siri this time. iOS 27, expected at WWDC June 8, will reportedly include a standalone Siri app with a chatbot interface powered by Gemini, capable of reading across your messages, email, and notes to take action inside third-party apps. Apple Intelligence launched with demos that outran the product by a wide margin. The June 8 keynote is worth watching, not believing in advance.
One Tool Worth Knowing
Claude Code auto mode, available now inside the Claude Desktop app.
If you already have a Claude Pro subscription, you have Claude Code. Open the Desktop app and click the Code tab at the top. That is the entry point, no terminal required.
The new auto mode means Claude can work through multi-step tasks without stopping to ask for approval at each action. You describe what you want in plain English, Claude builds or executes it, and you review the result.
A practical starting point: pick one task you do every week that is repetitive and low-stakes. Something like reformatting a CSV, pulling data from a report into a summary, or generating variations of copy from a brief. Describe it to Claude the way you would describe it to a colleague. See what comes back.
Austin Lau, who ran Anthropic’s entire performance marketing operation solo for nearly ten months using Claude Code, started with a calculator app. Not because it was useful. Because he needed proof that the thing worked before he trusted it with anything real. That is a good instinct. Start with zero stakes, build confidence, then bring it to the work that matters.
More on building with Claude Code: The Architecture of Building With AI.
“Unless the Lord builds the house, the builders labor in vain” (Psalm 127:1). Clarity about what you are making is not a productivity strategy. It is stewardship. The tools keep getting faster. The question that outlasts them all is whether you are building something that matters to God, or just something that ships.
Comments
Back to all posts