2026-03-31
By Vadym · Generated with AI, curated by me
The cleanup phase is here. OpenAI kills its most expensive vanity project, Reddit draws a line between humans and bots, and Shopify quietly turns every AI chatbot into a storefront. Meanwhile, a leaked model nobody was supposed to see is raising alarms in Washington.
OpenAI announced it is shutting down Sora, its AI video generation app and API. The app closes April 26; the API winds down September 24. The numbers are staggering: Sora was burning roughly $15 million per day in inference costs against just $2.1 million in total lifetime revenue. User count peaked at around one million before collapsing below 500,000. Disney, which had committed $1 billion to a Sora partnership, learned of the shutdown less than an hour before the public announcement — and the deal died with it. OpenAI says it will redirect the compute toward AGI research and robotics.
A misconfigured content management system exposed nearly 3,000 unpublished assets from a major AI lab, including details of a new model tier codenamed “Capybara” (internally called Mythos). The model sits above the current flagship tier and scores dramatically higher on coding, academic reasoning, and cybersecurity benchmarks. The lab is privately warning top government officials that Mythos makes large-scale cyberattacks significantly more feasible. The model is currently in restricted early-access testing and is described internally as a “step change” in capabilities.
Shopify flipped the switch on Agentic Storefronts, making every eligible merchant’s products discoverable inside ChatGPT, Microsoft Copilot, Google AI Mode, and the Gemini app — opt-out, not opt-in. The move came as OpenAI retreated from its “Instant Checkout” model that kept transactions inside the chat window and charged merchants 4%. The new design redirects buyers to the merchant’s own checkout instead. AI-driven traffic to Shopify stores is up 7x since January 2025, with AI-attributed orders up 11x over the same period.
Reddit’s new automated accounts policy goes live March 31. Bots built on Reddit’s developer platform can now display [App] labels on their profiles. Accounts flagged for suspicious automated behavior face on-device human verification via FaceID or passkeys — though Reddit stresses this won’t be a sitewide requirement. The company is removing over 100,000 bot accounts daily. AI-generated content itself isn’t banned; the policy targets accounts that aren’t run by humans, not humans who use AI tools to create content.
Orbital compute startup Starcloud raised $170 million at a $1.1 billion valuation, betting that the AI boom will eventually push data center economics beyond Earth. The company plans an 88,000-satellite constellation using near-continuous solar power in orbit to run AI workloads. Starcloud has already tested with NVIDIA hardware and AWS-related technology in prior launches. Reuters reports the raise reflects growing lender confidence that space-based AI infrastructure is bankable, not speculative.
Today’s stories share a theme: the industry is growing up, and growing up means making hard choices. OpenAI killed its most impressive demo because the economics didn’t work. Reddit is drawing lines between humans and machines because the bots got too good to ignore. Shopify is routing commerce through AI because that’s where the buyers already are. And Qodo exists because the code AI writes still needs a second pair of eyes. Even the wildest story — data centers in space — is driven by a mundane constraint: we’re running out of power on the ground. The common thread isn’t ambition, it’s accountability. The flash-and-demo era is giving way to the show-me-the-unit-economics era. The companies that survive this transition are the ones treating AI as engineering, not magic.
— Boba
Curated by Vadym