Sam Altman deleted TikTok because he could not stop scrolling. OpenAI killed Sora for the same reason.

TL;DR

Altman said he got “addicted” to TikTok while studying it for Sora. He deleted it after a 3-hour Saturday session. OpenAI killed Sora in March. The company says ChatGPT is designed not to trap users.

Sam Altman deleted TikTok because he could not stop using it. The OpenAI CEO said on the “Relentless” podcast that he forced himself to learn the app while OpenAI was building Sora, its AI-generated video social network. “I made myself get addicted to TikTok, because I wanted to just learn,” he said. He started with 10 minutes before bed. Then it was an hour. Then it was three hours on a Saturday afternoon. “I was really enjoying it in the moment like a drug,” he said. He deleted the app.

OpenAI deleted Sora too. The company scrapped the app in March after roughly six months, deciding that video generation consumed too much compute at a time when Anthropic and other competitors were pushing the frontier AI race further. Sora had attracted a Disney partnership and significant attention, but the economics did not work. The parallel is hard to miss: Altman found TikTok too consuming to keep, and OpenAI found Sora too consuming to sustain.

Altman discussed the experience in the context of the Jony Ive AI device, which OpenAI has said it will unveil before the end of the year. Altman has said he wants the device to cut through the notification noise that makes phones feel like “walking through Times Square.” Google rebuilt Search around AI summaries to keep users inside its ecosystem longer. Altman is arguing OpenAI should do the opposite: build a product people use briefly and leave. “Instead of measuring success by time spent or clicks, we care more about whether you leave the product having done what you came for,” OpenAI said last August.

The admission is revealing because it comes from the person building the AI systems that power the next generation of recommendation algorithms. Altman experienced firsthand that TikTok’s algorithm is, in his words, “too powerful,” and that even someone who understood exactly what was happening could not resist it. Wharton researchers found that people accept AI outputs 80% of the time, even when wrong. TikTok’s algorithm is not wrong. It is precisely right about what keeps you watching. The CEO of the world’s most valuable AI company could not outsmart it. That is the most honest product review TikTok has ever received.

Get the most important tech news in your inbox each week.