Summary

Sora’s shutdown looks like a clean product failure but a consequential warning about AI economics. The Sora team’s farewell promised timelines for the app, API, and preserving users’ work. Sora briefly reached the top of the App Store before proving to be a “novelty,” leaving Ben to call its closure a “very bloodless killing.” OpenAI’s product instincts were defensible; the standalone social network appears not to have reached critical mass.
Copyright controls may have removed the feature that gave Sora its best chance of surviving. Ben argued that early interest may have come from manipulating recognizable figures and characters; without that shared cultural material, interest may have fallen sharply. Andrew saw a plausible recurring meme tool—not an Instagram rival—and Ben called the unrealized version “the greatest meme maker of all time.”
Cloud video generation reverses the zero-marginal-cost conditions that helped earlier social products compound. Instagram began with filters computed locally and distribution handled by the App Store, then added infrastructure as its value grew. Sora incurred heavy compute costs from day one—Andrew cited an unverified “$15 million a day” estimate—creating an “all-or-nothing scenario” before OpenAI had figured out how to make money from it.
OpenAI’s enterprise turn may be economically necessary even as it leaves the consumer-platform thesis untested. Ben once envisioned a roughly billion-user consumer base plus advertising creating a revenue-and-R&D flywheel, but OpenAI was late to start monetizing and “we don’t know if that would’ve worked.” Enterprise customers will pay directly for productivity, Codex is compelling, and servers must be funded—though competition with Anthropic, Microsoft, and others will be “tooth and nail.”
Private-company revenue disclosures obscure the OpenAI–Anthropic race more than they resolve it. Ben said reported Anthropic ARR may resemble GMV by including total customer payments, while OpenAI’s numbers reflect the amount above what goes to cloud providers. Even so, “trajectories are real”: Anthropic’s appears more vertical, while OpenAI’s absolute lead might still be larger. Andrew said both companies could go public within the next couple quarters; Ben viewed that as public-company financing working as it should.

Deep dive

1. Sora went from App Store phenomenon to bloodless shutdown

The Sora team’s farewell promised timelines for the app, API, and preserving users’ work. Ben’s blunt question—“Did anything made with Sora matter?”—captured the gap between its launch frenzy and its eventual status as a novelty.
Ben noted that Sora had reached the top of the App Store before proving to be a novelty, while joking that Meta’s Vibes “won” despite his not opening it for months. His broader verdict: launching a standalone creative community was audacious because Facebook’s additional social-network launches generally failed.
Ben called the shutdown “a very bloodless killing,” reasoning that if no one was using the app, its disappearance would have little effect. He said Disney seemed to be the primary casualty, and Andrew noted its deal had also been rolled back. Andrew nevertheless credited OpenAI’s product instincts: Sora scared Hollywood and generated attention, but “the product itself isn’t really a business for them.”
Ben also raised the strategic question of whether Sora functionality should have been put into ChatGPT rather than launched as a standalone social product.

2. Copyright enforcement may have killed Sora’s strongest use case

Ben’s theory was that Sora’s early surge depended on users being able to animate recognizable figures—“Pikachu doing crazy things”—and that copyright crackdowns may have helped kill that interest. His larger thesis was that things people have in common will matter more and more; he pointed to live events as an extreme example of this shared cultural value.
Andrew thought people might still use Sora semi-regularly to enter The Matrix or remix familiar movie characters. Ben agreed that editing actual scenes could have made it “the greatest meme maker of all time.”
Andrew’s distinction is worth keeping: becoming a powerful meme tool was always more plausible than building a network rivaling Instagram, TikTok, or Twitter. Ben’s partial rebuttal was that heavy feature usage supplies the ingredients for a network, even if it never guarantees one.

3. Sora inverted Instagram’s low-cost path to network effects

Instagram began as an independently useful filter app: computation happened on the phone, Apple handled distribution, and users shared results into existing networks. It then followed the “come for the feature, stay for the network” progression, even lifting Twitter’s social graph wholesale.
Sora put community features in from the outset, yet most Sora videos people saw were probably viewed on other social networks. Ben wondered whether adding the social layer at launch was itself a mistake, while stressing that the vast majority of new social networks fail and joking that predicting every one will fail gives you a great track record.
The deeper problem was sequencing economics. Instagram’s marginal costs began “basically zero” and rose alongside its value; cloud-generated video starts expensive, before OpenAI had figured out monetization. Sora either became enormous or died within months—and even enormous usage would not have answered whether revenue exceeded costs.

4. OpenAI’s consumer flywheel was too late to test

Asked whether OpenAI’s enterprise pivot could coexist with consumer ambitions, Ben’s honest answer was, “I don’t know.” AI’s uncertainty—and especially its compute costs—makes assumptions imported from traditional software less reliable.
His prior thesis treated consumer as the ultimate prize: a defining company appears perhaps once every 15–20 years, with Facebook as the archetype. He envisioned a roughly billion-user consumer base supporting advertising, whose revenue would fund R&D and widen an “insurmountable lead.”
The frustration is that OpenAI was late to start monetizing, so “we don’t know if that would’ve worked.” Enterprise is now the clearest short-to-medium-term revenue source because customers pay for productivity and Codex is compelling, but there is no monopoly: OpenAI must fight Anthropic, Microsoft, and others “tooth and nail.”

5. Revenue definitions blur Anthropic’s apparent acceleration

Andrew contrasted OpenAI with Microsoft: Microsoft’s airtight enterprise monopoly taught consumers to use its products at work. OpenAI faces a fragmented enterprise market where several tools may coexist, and Anthropic’s share remains unknowable while both companies are private.
Ben said a story in The Information suggests Anthropic’s reported recurring revenue resembles GMV—the total customers pay—while OpenAI’s numbers reflect the amount above what goes to cloud providers. His Amazon analogy separated first-party retail revenue, which includes the item and its cost, from marketplace revenue, which records only Amazon’s skim.
Different methodologies do not erase momentum: “trajectories are real,” and Anthropic’s currently looks “more vertical,” even if OpenAI’s lead could be larger than reported comparisons imply. Andrew said both might go public within the next couple quarters; Ben welcomed that because the companies need money and said it represented the public-company concept working as it should.

The Ezra Klein Show: How Fast Will A.I. Agents Rip Through the Economy?68. [Bonus] Industrial AI Pitfall Guide: When 180 Years of Industrial Engineering Meets Large Models