标签： 人工智能

研究加速：OpenAI 内部视角

研究加速：OpenAI 内部视角

OpenAI 首次对外披露内部研发团队使用编程智能体的真实强度，数据显示 2026 年已成为智能体工程爆发之年，而研究人员人均 AI 算力支出的陡增节点，很可能指向了尚未正式命名的下一代模型。

“Cheating and explaining the answer [to your child] are closer than you’d think.” Here’s my new episode with @suekhim, co-founder of @brilliantorg, who showed me how AI can help kids think for themselves. We talked ab…

“Cheating and explaining the answer [to your child] are closer than you’d think.” Here’s my new episode with @suekhim, co-founder of @brilliantorg, who showed me how AI can help kids think for themselves. We talked ab...

Brilliant 联合创始人苏·基姆（Sue Khim）在播客访谈中提出，AI 教育产品的核心原则应是“永远不直接告诉学习者答案”。这一观点正在挑战当前 AI 辅导工具“即时给答案”的主流设计思路，引发对 AI 学习工具价值边界的讨论。

To calibrate you all on which reasoning effort to use for Astra, know that GPT-6 Astra on low performs better than GPT-5.6 Sol on high. If you were using high reasoning efforts with Sol and were happy, I suggest you m…

To calibrate you all on which reasoning effort to use for Astra, know that GPT-6 Astra on low performs better than GPT-5.6 Sol on high. If you were using high reasoning efforts with Sol and were happy, I suggest you m...

OpenAI 核心成员 Thibault Sottiaux 在 X 上发文，称 GPT-6 Astra 即便在低推理强度下，表现也优于 GPT-5.6 Sol 在高推理强度下的水平。这意味着 Astra 的模型基础能力大幅跃升，用户已不需要像使用旧模型那样依赖高推理模式来获取最佳答案。

Seattle Times和Newsday起诉OpenAI和Microsoft侵犯版权

Seattle Times和Newsday起诉OpenAI和Microsoft侵犯版权

《西雅图时报》与《新闻日报》联合起诉 OpenAI 和微软，指控其未经许可使用新闻报道训练 AI 模型，并要求销毁相关训练数据和模型。这是传统媒体与 AI 公司在版权问题上的又一次正面冲突，反映出内容授权纠纷正在从个案走向系统性对抗。

OpenAI says it hit its “automated research intern” goal, its researchers now use 3.1 agent-workdays per human workday, and top users spend $7,000+/day on tokens (OpenAI)

OpenAI says it hit its "automated research intern" goal, its researchers now use 3.1 agent-workdays per human workday, and top users spend $7,000+/day on tokens (OpenAI)

OpenAI 宣布其内部设定的“自动化研究实习生”目标已达成，研究团队如今每个人类工作日可调度约 3.1 个“Agent 工作日”，同时高端用户日均 Token 消耗费用已突破 7000 美元。这组数据意味着 AI Agent 从演示走向了真实的高强度研发协作场景。

OpenAI称已实现其打造自动化研究实习生的目标

OpenAI称已实现其打造自动化研究实习生的目标

OpenAI 于 2026 年 9 月宣布，已实现此前设定的“自动化研究实习生”目标，并称正在推进更高级别的“自动化 AI 研究员”，预计在 2028 年 3 月前完成。这一进展意味着 AI 已能独立完成部分需要数天时间的专项研究任务，但也伴随着系列安全事件带来的隐忧。

Analysis: since October, Anthropic has entered into agreements for at least 14.8 GW of compute capacity and may spend as much as $517B over the next decade (Valida Pau/The Information)

Analysis: since October, Anthropic has entered into agreements for at least 14.8 GW of compute capacity and may spend as much as $517B over the next decade (Valida Pau/The Information)

The Information 分析显示，Anthropic 自今年 10 月起已签署至少 14.8 吉瓦（GW）算力协议，未来十年相关支出可能高达 5170 亿美元。这笔巨额投入反映出头部 AI 实验室正在将算力锁定视为决定模型竞争力的核心战略。