OpenAI 缓存升级与用量修复OpenAI Cache Upgrade and Usage Fix
OpenAI Codex & ChatGPT 团队成员 Thibault Sottiaux 宣布 API 现支持 12 小时缓存,登录用户在 Codex 与 Work 中缓存永久有效,无回访惩罚,确保命中缓存。同时修复了约 4 万账户用量限制不一致的问题,将单独联系用户并补发额度。
Thibault Sottiaux of Codex & ChatGPT at OpenAI announced that the API now supports 12-hour caching, while logged-in users on Codex and Work keep cache forever with no penalty for returning to old conversations. Separately, about 40,000 accounts had inconsistent usage limits applied; the team is contacting them and adding credits.
Thibault Sottiaux of Codex & ChatGPT at OpenAI announced that the API now supports 12-hour caching, while logged-in users on Codex and Work keep cache forever with no penalty for returning to old conversations. Separately, about 40,000 accounts had inconsistent usage limits applied; the team is contacting them and adding credits.
OpenAI Codex & ChatGPT 团队成员 Thibault Sottiaux 宣布 API 现支持 12 小时缓存,登录用户在 Codex 与 Work 中缓存永久有效,无回访惩罚,确保命中缓存。同时修复了约 4 万账户用量限制不一致的问题,将单独联系用户并补发额度。
查看原文 →查看原文 →
自主 Agent 成为新形态Autonomous Agents as the New Form Factor
Box CEO Aaron Levie 指出,AI 正从 LLM wrapper 跃升为面向特定任务或领域的自主系统。更强的工具与计算机使用能力、垂直领域模型判断力提升,以及沙箱、文件系统等 harness 基础设施,使 agent 能可靠执行复杂工作流,并已出现在 IT 自动化、网络安全、药物发现等场景。
Box CEO Aaron Levie argues we now have an emerging form factor for AI: autonomous systems dedicated to specific tasks or domains. Better tool and computer use, improving model judgment in verticals, and maturing harnesses (sandboxes, file systems) let agents tackle sophisticated work with human oversight, already appearing in IT automation, cyber ops, drug discovery and more.
Box CEO Aaron Levie argues we now have an emerging form factor for AI: autonomous systems dedicated to specific tasks or domains. Better tool and computer use, improving model judgment in verticals, and maturing harnesses (sandboxes, file systems) let agents tackle sophisticated work with human oversight, already appearing in IT automation, cyber ops, drug discovery and more.
Box CEO Aaron Levie 指出,AI 正从 LLM wrapper 跃升为面向特定任务或领域的自主系统。更强的工具与计算机使用能力、垂直领域模型判断力提升,以及沙箱、文件系统等 harness 基础设施,使 agent 能可靠执行复杂工作流,并已出现在 IT 自动化、网络安全、药物发现等场景。
查看原文 →
Sam Altman 任命安全负责人Sam Altman on Superintelligence Safety Lead
OpenAI CEO Sam Altman 表示非常感激 Aleksander 接受邀请负责 superintelligence 的安全与安全挑战,这些挑战巨大,而他具备优秀的技能组合来应对。
OpenAI CEO Sam Altman expressed gratitude that Aleksander accepted the request to lead on the immense safety and security challenges of superintelligence, praising his skill set for the role.
OpenAI CEO Sam Altman expressed gratitude that Aleksander accepted the request to lead on the immense safety and security challenges of superintelligence, praising his skill set for the role.
OpenAI CEO Sam Altman 表示非常感激 Aleksander 接受邀请负责 superintelligence 的安全与安全挑战,这些挑战巨大,而他具备优秀的技能组合来应对。
查看原文 →
Google Gemini Trusted Tester 里程碑Google Gemini Trusted Tester Milestone
Google VP Josh Woodward 宣布 GeminiApp Trusted Tester 社区达到重要里程碑,Web、Android 和 iOS 上早期功能需求旺盛。已报名者全部开放,新用户可继续注册。他同时引用 Sundar 的话:"我们如此致力于走在前沿。这已融入我们的 DNA。"
Google VP Josh Woodward announced a major milestone for the GeminiApp Trusted Tester community, with strong demand for early features on Web, Android and iOS. Existing sign-ups are fully in; new ones remain open. He quoted Sundar: "We are so committed to being at the frontier. It’s in our DNA."
Google VP Josh Woodward announced a major milestone for the GeminiApp Trusted Tester community, with strong demand for early features on Web, Android and iOS. Existing sign-ups are fully in; new ones remain open. He quoted Sundar: "We are so committed to being at the frontier. It’s in our DNA."
Google VP Josh Woodward 宣布 GeminiApp Trusted Tester 社区达到重要里程碑,Web、Android 和 iOS 上早期功能需求旺盛。已报名者全部开放,新用户可继续注册。他同时引用 Sundar 的话:"我们如此致力于走在前沿。这已融入我们的 DNA。"
查看原文 →查看原文 →