OpenAI’s Sora wobble and product pivot
OpenAI’s Sora video tooling has been flaky — Sora 1 was retired and Sora 2’s text‑to‑video is frequently failing, with developers citing scaling, moderation and cost issues reported. At the same time leadership has signalled a companywide 'code red' to refocus on developer productivity and core apps reported, and OpenAI plans to embed Sora into ChatGPT to unify text and video workflows announced.
OpenAI’s support documentation lists Sora 1 as officially retired for U.S. users on March 13, 2026 (help.openai.com). An OpenAI developer-forum bug thread records Sora 2 video jobs remaining in “in_progress” at 99% with no completion for some users as early as October 15, 2025 (community.openai.com). Multiple public logs and community posts show recurring “We’re under heavy load” errors and capacity reports, with StatusGator aggregating 25 user-submitted Sora incident reports over a 24‑hour window on March 16, 2026 (statusgator.com). TechCrunch documented OpenAI disabling video generation for certain Sora users amid capacity constraints in March 2025, illustrating that the company has been managing load limits since at least last year (techcrunch.com). The Information reported that OpenAI plans to embed Sora’s video-generation capability directly into ChatGPT — a move The Information says could materially increase usage and costs — and OpenAI has not publicly confirmed the integration as of the report date. (theinformation.com) Industry outlets note the strategic motive: ChatGPT’s weekly reach was reported at roughly 900 million users in February 2026, and standalone Sora installs reportedly dropped about 45% in January 2026, factors The Information and several follow-ups cite as drivers for folding Sora into ChatGPT to boost engagement. (engadget.com)
