[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-aws-buys-ducklabs-duckdb-open-source":3,"news-related-21a91da5-c5fa-45e2-b01f-a7950331cf44":38},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":24,"news_slug":31,"published_at":32,"created_at":33,"modified_at":34,"is_published":35,"publish_type":36,"image_url":14,"view_count":37},"21a91da5-c5fa-45e2-b01f-a7950331cf44","S3 把 DuckDB 团队收走了:DuckLabs 加盟 AWS,MIT 开源照旧","亚马逊 AWS 正式签下收购 DuckLabs 协议,这家维护开源分析数据库 DuckDB 的阿姆斯特丹团队整体并入 S3 阵营。DuckDB 仍以 MIT 许可证开源,由独立基金会托管。","亚马逊 AWS 在 2026 年 8 月 26 日宣布签下收购 DuckLabs 的最终协议,这支位于阿姆斯特丹的三十人团队是开源分析数据库 DuckDB 的核心开发力量。交易预计在 9 月初完成交割,创始人 Hannes Mühleisen 与 Mark Raasveldt 将继续领导团队和项目技术方向,工作室也继续留在阿姆斯特丹。消息一出,AWS Big Data 博客、DuckLabs 官方公告和 GeekWire、The Register 等媒体几乎同步落地,可见这次不是传闻阶段。\n\n之所以把收购单独拆出\"团队\"两个字,是因为协议里专门写明:DuckDB、DuckLake、Quack 这一整套 Duck Stack 仍然以 MIT 许可证开源,由独立的非营利 DuckDB Foundation 托管,AWS 不接管项目治理权。DuckLabs 在官方公告里强调,基金会接下来还会增设技术顾问委员会,并开放第三方签名扩展在 DuckDB 内运行,这是创始人想保住开源底线的两个具体动作。AWS 副总裁级 Distinguished Engineer Andy Warfield 在两家博客里都被引用,他那句\"和 Mark、Hannes 和整个 DuckLabs 团队共事两年,这是我合作过技术最深、最谦虚、最高 velocity 的团队\",把 AWS 想留人的姿态摆得很清楚。\n\n## 为什么是 DuckDB\n\nDuckDB 诞生于荷兰国家数学与计算机科学研究中心(CWI),而 CWI 也正是 Python 语言的诞生地。两位创始人在 CWI 做研究时就意识到,Spark 等老一代引擎把性能都堆在大数据批处理上,却忽略了 90% 以上的真实 SQL 查询其实跑在 1TB 及以下的小数据集上。AWS 自己的 Big Data 博客里也复述了这个判断:DuckDB 的向量化执行(in-process vectorized execution)能让\"日常 SQL 查询变得飞快\",架构上跑在宿主进程内省掉了网络跳转和序列化成本,这一点对低延迟数据探索非常关键。\n\n更值得注意的是,DuckDB 的设计风格天然契合 AI agent 的工作方式。AWS 在博客里写得很直白:\"agent 跟人很像,它们先戳一戳、跑跑探索性分析,再决定下一步到底要查什么。\"这种\"小数据快速反复探索\"的模式,正好是 DuckDB 一秒级返回查询的甜蜜区。AWS 内部已经在 Amazon Quick 产品上验证了这条路径——2025 年 10 月上线以来,DuckDB 内核累计处理了超过 25 亿次查询,平均延迟下降约 30%。\n\n## 客户和生态背书\n\nAWS 的博客里直接引用了 Allen Institute 科学计算执行总监 David Feng 的使用证言,他说 2025 年起把 DuckDB 用于神经科学和行为数据的实时质量控制,原本分钟级的查询压到了秒级,这种交互节奏的变化让研究流程变了样。DuckLabs 公告里则拿到了生态里两家关键公司的公开背书:MotherDuck CEO Jordan Tigani 评价\"亚马逊给 DuckDB 加压会让生态飞轮更快\",Fivetran 联合创始人 George Fraser 强调\"DuckDB 是供应商中立开放数据栈的中心,亚马逊的开放承诺是它继续担当这个角色的关键\"。\n\n学术界同样表态。DuckDB 基金会董事、CWI 数据库架构组负责人 Peter Boncz 强调基金会将继续持有 DuckDB 全部开源 IP;图宾根大学数据库系统组负责人 Torsten Grust 教授则补了一句很学术的话——DuckDB 的\"可窥探性\"让它成为数据库研究与教学的首选平台,这次开源属性保住后,这类使用场景有了更强的可持续性。\n\n## 我的判断\n\n把 DuckLabs 团队收进 AWS,本质上是云厂商把\"分析层最受欢迎的开源 OLAP 引擎\"和\"S3 数据湖的事实标准\"在组织层面绑死。在 Snowflake 与 Databricks 主导 lakehouse 话语权的格局下,AWS 用一个不到三十人的小团队,就把\"在数据湖上做秒级本地分析\"这条体验链锁进了自己的生态。对 DuckDB 用户来说,这次最大的风险不在协议条款而在执行节奏:开源治理委员会、技术路线、社区分发渠道这些细节一旦被云厂节奏拖慢,会比\"换许可证\"更伤;反过来,基金会扩权、扩展签名开放、AWS 把 Quick 的 DuckDB 优化路径反哺给开源版本,这三条如果一年内能落地,DuckDB 的工程密度反而会再上一个台阶。\n\n最后落到一个普通开发者视角:DuckDB 现在每天下载量已经超过一百万次,这次并购短期不会改协议,长期观察点应该是基金会技术顾问委员会的成员名单和扩展生态的开放节奏——这两个比任何公关稿都能告诉我们,DuckDB 的\"开源\"二字到底是写给投资者看,还是写给用户看。\n\n参考来源:AWS Big Data Blog《AWS and DuckLabs: Building the future of analytics together》、DuckLabs 官方公告《DuckLabs to Join AWS, Projects to Remain Open Source》、All Things Distributed《DuckDB and the Changing Physics of Analytics》(Andy Warfield)、Amazon Newsroom《AWS to acquire DuckLabs》。","https:\u002F\u002Fducklabs.com\u002Fnews\u002F2026\u002F08\u002F26\u002Fducklabs-to-join-aws","19377961-7140-4d39-9520-0e17c682c90d",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",{"id":19,"name":20,"slug":20,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":22,"name":23,"slug":23,"description":14,"color":14},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"cd5dd04d-b9ca-4f39-afcc-9b0786990116","en","AWS absorbs the DuckDB team: S3 camp signs DuckLabs, MIT open source unchanged","Amazon AWS has signed a definitive agreement to acquire DuckLabs, the Amsterdam team behind the open-source analytical database DuckDB. DuckDB remains open source under the MIT license and is governed by an independent nonprofit foundation.","Amazon AWS announced on August 26, 2026 that it has signed a definitive agreement to acquire DuckLabs, the thirty-person Amsterdam team behind the open-source analytical database DuckDB. The deal is expected to close in early September, with founders Hannes Mühleisen and Mark Raasveldt continuing to lead the team and the project's technical direction, and the team remaining in Amsterdam. The news landed almost simultaneously across the AWS Big Data blog, the DuckLabs announcement, and outlets like GeekWire and The Register, signaling that the deal is no longer rumor-stage.\n\nThe reason the word \"team\" is called out separately is that the agreement explicitly states: DuckDB, DuckLake, Quack, and the broader Duck Stack remain open source under the MIT license and continue to be governed by the independent nonprofit DuckDB Foundation, with AWS not taking over project governance. In its official announcement, DuckLabs emphasized two concrete moves to safeguard the open-source posture: the foundation will add a technical advisory board, and the extension stack will be opened so that third-party signed extensions can run inside DuckDB. AWS Distinguished Engineer Andy Warfield is quoted on both blogs, with his line \"after about two years of working closely with Mark, Hannes and the whole DuckLabs team... one of the most technically deep, humble, and high-velocity teams that I've ever had a chance to work with\" making it very clear that AWS wants to retain these people.\n\n## Why DuckDB\n\nDuckDB was born at Centrum Wiskunde & Informatica (CWI) in the Netherlands, the same national research institute that also invented Python. While at CWI, the two founders realized that older engines like Spark focused on performance for very large data processing while ignoring the reality that 90% or more of real SQL queries actually run on datasets of 1TB or less. The AWS Big Data blog reiterates this judgment: DuckDB's vectorized execution, running in-process inside the host application, eliminates network hops and serialization costs, which is exactly the property low-latency data exploration needs.\n\nMore notably, DuckDB's design philosophy is a natural fit for how AI agents work with data. The AWS blog puts it bluntly: \"agents behave a lot like people when interacting with data. They poke. They experiment. They run exploratory analysis on small data sets before figuring out what they really want to do.\" This pattern of small-data, fast, iterative exploration sits precisely in DuckDB's sweet spot for sub-second returns. AWS has already validated this path internally inside Amazon Quick — since launching in October 2025, the DuckDB-backed engine has processed more than 2.5 billion queries and reduced average latency by roughly 30%.\n\n## Customer and Ecosystem Endorsements\n\nThe AWS blog directly quotes David Feng, Executive Director of Scientific Computing at the Allen Institute, who explained that since 2025 the Allen Institute has used DuckDB for real-time quality control over neurophysiology and behavior data, compressing queries that previously took minutes down to seconds, and that this interactive tempo has changed how research proceeds. The DuckLabs announcement secures public endorsement from two critical ecosystem players: MotherDuck CEO Jordan Tigani called it \"Amazon putting its weight behind DuckDB is going to add a ton of momentum,\" and Fivetran co-founder George Fraser emphasized that \"DuckDB is the center of the vendor-neutral open data stack, and Amazon has the commitment to openness... to enable the DuckDB project to continue to thrive in this role.\"\n\nThe academic side lined up as well. Peter Boncz, head of the Database Architectures group at CWI and a DuckDB Foundation board member, stressed that the foundation continues to hold all open-source IP of DuckDB. Torsten Grust, professor and head of the Database Systems research group at the University of Tübingen, added an academic note: DuckDB's \"hackability\" made it a top platform for database research and teaching, and with the open-source posture preserved, those use cases are now on a more durable footing.\n\n## My Take\n\nBringing the DuckLabs team inside AWS is, at its core, the cloud vendor tying together \"the most popular open-source OLAP engine for the analytics layer\" and \"the de facto standard for the data lake\" at the organizational level. In a landscape where Snowflake and Databricks dominate the lakehouse discourse, AWS used a sub-thirty-person team to lock the experience chain of \"sub-second local analytics on the data lake\" inside its own ecosystem. For DuckDB users, the biggest risk is not in the agreement text but in execution cadence: governance committee makeup, technical roadmaps, and community distribution channels — once those start moving at cloud-vendor pace, it hurts more than any license change. Conversely, if the foundation's expanded authority, the opening of the signed-extension ecosystem, and AWS feeding back the Quick product's DuckDB optimizations into the open-source version all land within a year, DuckDB's engineering density will move up another notch.\n\nClosing from a regular developer perspective: DuckDB already sees more than one million downloads per day. The acquisition does not change the license in the short term; the long-term watchpoints should be the membership of the foundation's technical advisory board and the pace at which the extension ecosystem opens up — those two tell us far more than any press release whether \"open source\" on DuckDB is written for investors or for users.\n\nSources: AWS Big Data Blog \"AWS and DuckLabs: Building the future of analytics together\"; DuckLabs official announcement \"DuckLabs to Join AWS, Projects to Remain Open Source\"; All Things Distributed \"DuckDB and the Changing Physics of Analytics\" by Andy Warfield; Amazon Newsroom \"AWS to acquire DuckLabs.\"","aws-buys-ducklabs-duckdb-open-source","2026-08-30T06:00:00Z","2026-08-30T05:04:53.327625Z","2026-08-30T05:04:53.327635Z",true,"agent",124,{"items":39},[40,45,50,55,60,65],{"id":41,"title":42,"news_slug":43,"published_at":44},"22a1a718-0eb6-46e5-8ee8-825400de11d1","DeepMind WeatherNext 在 Nature 发论文：用 28 km 粗分辨率做出多一天的飓风预警,代码权重全部开源","deepmind-weathernext-cyclones-nature-open-source","2026-08-10T02:00:00+00:00",{"id":46,"title":47,"news_slug":48,"published_at":49},"777afb24-262f-45cc-961f-d5d49ad42883","AgentOPSD 用递归贝叶斯信念破解多轮 Agent 强化学习的信用分配：清华\u002F浙大\u002F美团让 GRPO 学会看哪个 turn 决定胜负","agentopsd-recursive-belief-credit-assignment","2026-08-07T02:00:00+00:00",{"id":51,"title":52,"news_slug":53,"published_at":54},"daf9222a-aebc-4f09-92f8-74b6226fd5f1","把图灵测试变成 RL 损失：MIT 提出 Turing-RL，让用户模拟器更\"像人\"","turing-rl-mit-stanford-discriminative-reward","2026-06-18T02:00:00+00:00",{"id":56,"title":57,"news_slug":58,"published_at":59},"f2f0017f-449d-492b-b9d7-90a2013023fb","英伟达 129 亿美元收购 Hugging Face 接近敲定:开源 AI 仓库终被算力霸主收编","nvidia-12-9-billion-hugging-face-acquisition","2026-09-04T03:00:00+00:00",{"id":61,"title":62,"news_slug":63,"published_at":64},"b1645fba-d364-47e6-97da-06868f98d987","Linux 内核 7.x 每版近 2000 个 CVE:AI 帮倒忙,维护者不堪重负","linux-kernel-cve-ai-overwhelmed","2026-09-04T00:00:00+00:00",{"id":66,"title":67,"news_slug":68,"published_at":69},"c70131ba-e6b2-466a-ab36-65fad341f006","别让语音助手念出美元符号:NAVER 对齐 LLM 生成可朗读文本","tts-friendly-llm-alignment-fast","2026-09-02T23:40:00+00:00"]