[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f6z5VqXvEwGeqBVsDzUI3HxlhmhVcVhYwvNZt1nkDuQM":3},{"code":4,"msg":5,"data":6},200,"操作成功",{"id":7,"title":8,"content":9,"digest":10,"source":10,"coverPath":11,"thumbsCoverPath":12,"isTop":13,"isShow":14,"baseClick":13,"clickCount":15,"createTime":16,"typeId":17,"isNewest":18,"newsInfoTypeRespVo":19,"voiceUrl":22,"voiceSize":23,"taskId":24,"releaseTime":25,"titleEn":26,"contentEn":27,"voiceUrlEn":28,"taskIdEn":29,"voiceSizeEn":30},1488,"亚马逊云科技推出三款AI智能体，Kiro可连续自主编程数日","\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\">IT之家12月3日消息，亚马逊云科技（AWS）于当地时间周二宣布推出三款新的AI智能体，官方称之为“前沿智能体”（frontier agents），其中一款能够学习用户的工作偏好，并可连续自主运行数日。\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"undefined\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F12\u002Fc3efc31ffea9445f8da72d1cd2b650fb\u002F123.png\" width=\"705\" height=\"undefined\" style=\"\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">据IT之家了解，这些智能体各自承担不同任务，包括编写代码、执行安全流程（如代码审查），以及自动化DevOps任务（例如在新代码上线时预防事故）。目前，这些智能体的预览版本已推出。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">AWS宣称，其名为“Kiro自主智能体”（Kiro autonomous agent）的前沿智能体能够连续自主工作数日之久。Kiro是一款基于AWS今年7月发布的同名AI编码工具构建的软件编码智能体。此前的Kiro工具主要用于“氛围编码”（vibe coding），但其目标是生成可投入生产的运营级代码。为确保代码可靠性，该AI必须遵循企业自身的软件编码规范。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Kiro通过一种名为“规范驱动开发”（spec-driven development）的理念实现这一点。在编码过程中，Kiro 会要求人类用户对其假设进行指导、确认或修正，从而逐步形成明确的开发规范。而Kiro自主智能体则通过扫描现有代码等多种训练方式，观察团队在各类工具中的实际工作方式。AWS表示，此后它便能独立开展工作。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">AWS首席执行官马特・加曼（Matt Garman）在AWS re:Invent大会主题演讲中介绍该产品时表示：“你只需从待办事项列表中分配一个复杂任务，它就能自主规划并完成这项工作。”他补充道：“它实际上会学习你偏好的工作方式，并随着时间推移不断加深对你代码、产品以及团队所遵循标准的理解。”\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">亚马逊称，Kiro具备“跨会话的持久上下文”（persistent context across sessions）能力。换言之，它不会因内存限制而遗忘任务目标。因此，亚马逊承诺，Kiro可在极少人工干预的情况下，独立处理任务数小时乃至数日。加曼举例说明：比如更新企业15个软件所使用的一小段关键代码，传统做法需逐一指派并验证每处更新；而如今只需向Kiro发出一条指令，即可一次性完成全部15处修复。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">为实现编码任务的端到端自动化，AWS还开发了“AWS安全智能体”（AWS Security Agent），该智能体可独立识别编码过程中的安全问题，在代码编写完成后进行测试，并提供修复建议。第三款智能体“DevOps 智能体”（DevOps Agent）则负责自动测试新代码的性能问题，或检查其与现有软件、硬件及云环境的兼容性。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">当然，AWS并非首家宣称支持长时间连续运行的AI智能体厂商。例如，OpenAI 上月就表示，其智能体型编码模型 GPT5.1 Codex Max同样设计用于长时间运行，最长可达24小时。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">不过，业界普遍认为，当前阻碍AI智能体广泛采用的最大障碍或许并非上下文窗口（即持续工作而不中断的能力）。大型语言模型（LLMs）仍存在幻觉和准确性问题，导致开发者不得不像“保姆”一样频繁监督。因此，许多开发者倾向于分配短小任务，并在推进前快速验证结果。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">尽管如此，在AI智能体真正成为可靠“同事”之前，扩大上下文窗口仍是必要前提。亚马逊此次发布的技术，无疑是朝着这一方向迈出的重要一步。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(136, 136, 136);\">【新闻来源】IT之家\u003C\u002Fspan>\u003Ca href=\" https:\u002F\u002Ffinance.sina.com.cn\u002Ftech\u002Fdigi\u002F2025-12-03\u002Fdoc-infznkss9779353.shtml\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(136, 136, 136);\"> \u003Cu>https:\u002F\u002Ffinance.sina.com.cn\u002Ftech\u002Fdigi\u002F2025-12-03\u002Fdoc-infznkss9779353.shtml\u003C\u002Fu>\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(136, 136, 136);\">（本网转发此文章，旨在为读者提供更多的信息资讯，所涉内容不构成投资、消费建议。文章事实如有疑问，请与有关方核实，文章观点非本网观点，仅供读者参考。）\u003C\u002Fspan>\u003C\u002Fp>","","https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F12\u002F834a7e2b9ed14bf2a15cca0e78dd2465\u002Fd3ceb214-e320-4d27-b9ff-62bd5ab11fa0.jpg","https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F12\u002Fthumbs\u002F834a7e2b9ed14bf2a15cca0e78dd2465\u002Fd3ceb214-e320-4d27-b9ff-62bd5ab11fa0.jpg",0,1,47,"2025-12-05 18:25",2,false,{"id":17,"name":20,"enName":21},"芯位视野","Xinwei Vision","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3A64de3d5b-1d03-4367-8488-5bb1f94ba2a6%3A0.wav?Expires=1765160321&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=RWEnpCPb9HBuIDOMQmRou%2B0clNQ%3D",6848266,"64de3d5b-1d03-4367-8488-5bb1f94ba2a6","2025-12-05 18:05","AWS launches three AI agents, Kiro can autonomously program for several days.","\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\">IT之家 December 3rd report, Amazon Web Services (AWS) announced on Tuesday local time the launch of three new AI agents, officially called \"frontier agents,\" one of which can learn users' work preferences and run autonomously for several days.\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"undefined\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F12\u002Fc3efc31ffea9445f8da72d1cd2b650fb\u002F123.png\" width=\"705\" height=\"undefined\" style=\"\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">According to IT Home, these agents each undertake different tasks, including writing code, performing security processes (such as code reviews), and automating DevOps tasks (for example, preventing incidents when new code is deployed). The preview versions of these agents are now available.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">AWS claims that its frontier agent named \"Kiro Autonomous Agent\" can work autonomously for several days. Kiro is a software coding agent built upon the same-named AI coding tool AWS released in July this year. The previous Kiro tool was mainly used for \"vibe coding,\" but its goal is to generate production-ready operational code. To ensure code reliability, this AI must follow the company's own software coding standards.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Kiro achieves this through the concept of \"spec-driven development.\" During the coding process, Kiro will ask human users to guide, confirm or correct its assumptions, gradually forming clear development specifications. The Kiro autonomous agent observes how teams actually work in various tools through multiple training methods, such as scanning existing code. AWS says it can then work independently.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">During his keynote speech at the AWS re:Invent conference, AWS CEO Matt Garman introduced the product, saying, \"You just need to assign a complex task from your to-do list, and it can autonomously plan and complete the work.\" He added, \"It actually learns your preferred way of working and gradually deepens its understanding of your code, products, and the standards your team follows over time.\"\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Amazon states that Kiro has the ability to maintain persistent context across sessions. In other words, it does not forget the task objectives due to memory limitations. Therefore, Amazon promises that Kiro can handle tasks for hours or even days with minimal human intervention. Garman gave an example: for instance, updating a small piece of critical code used by 15 enterprise software applications. Traditionally, each update needs to be assigned and verified individually. Now, you just need to give Kiro one instruction, and it can complete all 15 fixes at once.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">To achieve end-to-end automation of coding tasks, AWS also developed the \"AWS Security Agent,\" which can independently identify security issues during the coding process, test after the code is written, and provide repair suggestions. The third agent, the \"DevOps Agent,\" is responsible for automatically testing performance issues in new code or checking its compatibility with existing software, hardware, and cloud environments.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Of course, AWS is not the first company to claim support for AI agents that can run continuously for long periods. For example, OpenAI said last month that its intelligent coding model GPT5.1 Codex Max is also designed for long-term operation, up to 24 hours.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">However, the industry generally believes that the biggest obstacle to the widespread adoption of AI agents may not be the context window (the ability to work continuously without interruption). Large language models (LLMs) still have hallucination and accuracy issues, causing developers to frequently supervise them like a babysitter. Therefore, many developers prefer to assign small tasks and quickly verify the results before proceeding.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Despite this, expanding the context window remains a necessary prerequisite before AI agents truly become reliable \"colleagues.\" Amazon's recent technological release is undoubtedly an important step in this direction.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(136, 136, 136);\">[News Source] IT Home\u003C\u002Fspan>\u003Ca href=\" https:\u002F\u002Ffinance.sina.com.cn\u002Ftech\u002Fdigi\u002F2025-12-03\u002Fdoc-infznkss9779353.shtml\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(136, 136, 136);\"> \u003Cu>https:\u002F\u002Ffinance.sina.com.cn\u002Ftech\u002Fdigi\u002F2025-12-03\u002Fdoc-infznkss9779353.shtml\u003C\u002Fu>\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(136, 136, 136);\"> (This article is reposted by this website to provide readers with more information and news. The content involved does not constitute investment or consumption advice. If there are any questions about the facts of the article, please verify with the relevant parties. The views expressed in the article are not the views of this website, and are for reference only.)\u003C\u002Fspan>\u003C\u002Fp>","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3A9534d143-e4b8-4de1-abda-c1233ce23bab%3A0.wav?Expires=1774838447&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=LhU6u%2Bh0ioNNHwbDgMaRFjlUd4A%3D","9534d143-e4b8-4de1-abda-c1233ce23bab",7789404]