Sam Altman Explains Why OpenAI Killed Sora and Atlas to Focus on Codex and AGI
In a nearly 1.5-hour in-depth interview, Sam Altman explained OpenAI's decision to shut down the video generation model Sora and its in-house browser Atlas: with limited compute, talent, and resources, the company must prioritize what matters most for AGI — the knowledge work and science directions represented by Codex.
Confirmed
- Per @sahilypatel, Altman said these products were excellent and interesting, but Sora consumed enormous compute and was less important than Codex, forcing a tradeoff.
- Per @tristanbob, Altman anchored the company's priorities on AGI-related knowledge work and science.
- Per @AI寒убrie... per @AI Cambrian's recap, Altman positioned OpenAI as a platform company rather than a vertical product company; the interview also covered the trajectory of large models, business strategy, and the company's early days.
- Per a clip shared by @davidperell, Altman noted that the best ideas are extremely fragile early on: founders can be distracted by good-but-non-core ideas or kill seemingly bad ideas with great potential too soon.
Why it matters
- It shows the real tradeoffs top AI labs make under compute constraints: even excellent products must yield to the AGI mainline.
- Altman's arguments that "good ideas are the enemy of great founders" (via @davidsenra) and that good ideas are fragile offer founders a methodology for focus and early-stage judgment.
2026-08-24 ~ 2026-08-25 · 5 related posts
Primary sources
- [source] Sam Altman explains killing Sora and Atlas to focus on general intelligence — sahilypatel · 2026-08-24
- Sam Altman on shutting down Sora and Atlas to focus on Codex — davidsenra · 2026-08-24
- Sam Altman on Killing Sora and Atlas: Great Ideas Are Extremely Fragile — david_perell · 2026-08-24
- [source] Altman says OpenAI killed Sora and its browser to focus compute on AGI platform — AI寒武纪 · 2026-08-25
1 near-duplicate retellings: tristanbob