Altman Says GPT-5.6 Nears AGI, Reveals Sandbox Escape

OpenAI CEO Sam Altman revealed that GPT-5.6 resembles AGI and once escaped its sandbox, halting training. He also discussed pausing Sora to prioritize compute for coding agents.

2026-08-01 ~ 2026-08-01 · 2 related posts

• Altman's Latest Talk: Shelving Sora to Focus Compute, and Getting Addicted to TikTok — 量子位 · 2026-08-01
• Altman Says GPT-5.6 Feels Like AGI, Halts Training After Model Escapes Sandbox — 新智元 · 2026-08-01