OpenAI

Latest AI news, models and releases from OpenAI. ['API', 'Astra', 'Atlas', 'autonomous agent built on OpenAI models', 'autonomous model', 'ChatGPT', 'ChatGPT 4.0', 'ChatGPT 5.6', 'ChatGPT Codex', 'ChatGPT Enterprise', 'ChatGPT Go', 'ChatGPT-Live', 'ChatGPT Plus', 'ChatGPT-User', 'ChatGPT Work', 'Claude Code', 'Claude Fable 5', 'CLIP', 'Codex', 'Codex CLI', 'Codex Micro', 'Codex Security', 'Codex Security CLI', 'DALL-E', 'DALL-E 2', 'DALL-E 3', 'DALL·E 3', 'Daybreak', 'experimental prototype', 'Frontier', 'GPT', 'GPT-2', 'GPT2', 'GPT-2 XL', 'GPT-3', 'GPT 3.5', 'GPT-3.5', 'gpt-3.5-turbo', 'GPT-3.5-turbo', 'GPT-3.5 Turbo']

Research 🇷🇺

GPT-5.6 helps disprove the Maxwell conjecture — but not the one everyone thought

A preprint on arXiv claims to refute the so-called Maxwell conjecture about the maximum number of equilibrium points for point charges, with the key idea attributed to OpenAI's GPT-5.6 Sol model. The conjecture, formulated in 2007, concerned a bound of (n−1)² non-degenerate equilibrium points for n charges. The counterexample uses a configuration of five charges yielding at least 24 equilibrium points, exceeding the predicted 16.

OpenAIOpenAI
Habr — хаб ИИ03.08 · 13:02
Fresh news