哈工大、鹏城实验室、新加坡国立、复旦、北大 联合发布了一篇重磅综述 《AI Meets Brain: A Unified Survey on Memory System from Cognitive Neuroscience to Autonomous Agents》， 首次打破认知神经科学与人工智能之间的学科壁垒，系统性地将人脑记忆机制与 Agents 记忆统一审视，为设计真正「类人」的 Agent 记 忆系统奠定理论基石。 全文横跨认知神经科学与人工智能两大领域，涉猎相关文献共 400 篇。 跨学科突破：神经科学如何让 Agent 拥有「人类式」记忆？ 你是否想过 Agent 能像人类一样积累经验、不断成长？如今，这一愿景正加速走向现实。但是，现有研究要么只聚焦 AI 技术本身，要么对人脑记忆机制的借鉴浮 于表面，两个学科之间始终缺少真正的灵感碰撞。 什么是记忆？ 综述重新定义了记忆。记忆不仅仅是数据的存储，它也是认知的纽带。综述从认知神经科学到 Agent 对记忆进行了剖析： 1.认知神经科学角度：连接过去与未来的桥梁 在人脑中， 记忆不仅仅是回放信息，其本质是大脑存储和管理信息的过程。 记忆是连接过去 ...

在硅谷，围绕一个问题的争论正在升温： AI，尤其是 Agent，会不会取代 SaaS？ 最早给出明确判断的是SaaS 领域的知名专栏作者 Jamin Ball。 在《Long Live Systems of Record》一文中，他直言不讳地反对"Agent 会杀死一切旧系统"的说法。 她在最新文章《人工智能的万亿美元机遇：上下文图谱》中指出，传统系统的盲区不在于"数据"，而在 于"上下文"匮乏。 企业真实的运行逻辑，往往不记录在CRM 的标准化表格里，而是藏在例外的特批、临时的调整、跨部 门的 Slack 沟通中。 Gupta 将这些隐性的过程定义为「决策轨迹」。 当这些决策轨迹被持续记录，并在时间和业务对象之间连接起来，就会形成一种新的结构——上下文图 谱。 这不仅是数据的堆砌，更是对企业"推理过程"的复刻。下一个万亿级平台的机会，不是给旧系统装上 AI，而在于谁能抓住这些"数据"与"行动"之间的灰色地带。这才是AI创业公司需要抓住的真正机会。 今天，我们就来拆解这个超级赛道的核心逻辑。 01 上下文图谱：AI 时代企业最值钱的"第二资产" 上一代企业软件通过成为"记录系统"（Systems of Rec ...

中信证券：2026年看好Agent、多模态、情感陪伴及AI硬件的非线性增长 展望2026年，AI叙事将向纵深演绎，持续主导科技板块价值重塑。预计市场焦点将实现从"模型迭 代"到"场景落地"的范式转移。算力作为基础设施仍是坚实的压舱石，硬件层将迎来"自主可控"与"景气 外溢"的双重共振：对内看好国产算力与半导体设备在自主可控趋势下的系统级突围；对外把握AI PCB 与存储在全球需求共振下的超级景气周期。同时，超额收益的阿尔法或将源于应用层的爆发：重点看好 Agent、多模态、AI+办公/Coding、情感陪伴及AI硬件的非线性增长。此外，头部模型与应用独角兽的 资本化进程，将为板块提供关键的估值锚点与重估契机。海外市场方面，尽管美股科技或面临波动，但 在业绩支撑下，20%的指数级收益仍是可预期的合理目标。 国泰海通：先进制造领域重点关注盈利能力突出、全球化布局深入的龙头价值重估 国内先进制造产业发展相对成熟，凭借完备的工业体系和显著的效率成本优势，已在全球范围内建立起 稳固的竞争力。新能源中，锂电规模与盈利能力全球领先，头部企业估值普遍低于海外龙头，性价比优 势突出；风电盈利能力虽弱于海外，但估值同样较低；高端 ...

移动支付网消息（慕楚）：2025年已经过去，对于全球支付产业来说，这一年的变化可谓翻天覆地。藉 此辞旧迎新之际，我们来盘点下2025年全球支付产业发展的10条明线与暗线。 1、稳定币成为全球支付关键基础设施 稳定币无疑是2025年全球支付产业的最大热词，《香港稳定币条例》、美国《天才法案》、欧洲 MiCA、Circle上市等热门事件，让稳定币迅速出圈，获得普罗大众的认可。然而新兴事物的发展必然带 来诸多争议，如弱势币种的抵制、系统性风险的考虑、传销与诈骗的盛行、反洗钱压力陡增等等话题， 但随着全球各大金融机构与支付巨头的认可，稳定币毫无疑问在2025年成为了全球支付关键基础设施。 2、全球CBDC发展的踌躇 在稳定币的全球强势兴起下，号称"官方稳定币"的CBDC大多踌躇不前。美国通过《天才法案》的同 时，还通过了《反CBDC监控国家法案》，全面禁止CBDC的发行；英国在7月考虑搁置数字英镑计划， 到年底则向企业征求数字英镑的价值；欧盟一直加速数字欧元的落地，但时间推迟到2029年；韩国在年 中暂停CBDC的二阶段试点，但经历韩元稳定币发展不利后，在12月再次启动该测试。 而我国在2025年上半年便开始准备将数 ...

Core Insights - The article emphasizes the transformative impact of AI on the cloud market, highlighting that AI is becoming a core growth driver for cloud vendors as more industries recognize its potential to optimize business efficiency and create value [3][4][6] - Cloud vendors are shifting from being infrastructure providers to AI capability providers, with the development of Agents seen as a method to unlock the value of Tokens [6][11] - There is a growing demand for reliable and low-barrier Agent development platforms to help enterprises transition from concept validation to large-scale application [3][4][6] Group 1: Market Growth and Opportunities - AI-driven growth has created new business opportunities for cloud vendors, with Alibaba Cloud's AI-related revenue experiencing triple-digit year-on-year growth for nine consecutive months [1] - AWS anticipates that future Token revenue from MaaS platforms will be comparable to its EC2 computing product revenue [1] - Google Cloud's annual revenue exceeds $50 billion, largely driven by AI [1][3] Group 2: Agent Development and Implementation - The article discusses the need for a comprehensive framework for Agent development, which includes tools for model customization, Agent development, operation, and security governance [3][4][7] - Companies like Firefly Engine and AWS are actively providing solutions to lower the barriers for Agent development, emphasizing ease of use and security [7][9] - The development of Agents is expected to lead to significant efficiency improvements, as evidenced by case studies where projects were completed in a fraction of the time previously required [4][6] Group 3: Challenges in Agent Adoption - Despite the optimistic outlook, many enterprises face significant obstacles when moving from concept validation to production, with 93% of customers encountering major challenges [6] - The challenges stem from data and engineering aspects, where the requirements for large-scale applications differ significantly from those in the concept validation phase [6][11] Group 4: Strategic Shifts Among Cloud Vendors - Cloud vendors are increasingly focusing on developing foundational models that are tailored for Agent development, with companies like Tencent and Baidu making organizational adjustments to enhance their model capabilities [11][13] - The shift towards MaaS business models is becoming a core metric for cloud vendors, with Firefly Engine viewing AI Token usage as a key performance indicator [13] - Alibaba Cloud aims to position itself as a leading player in the AI era, with a focus on comprehensive AI cloud solutions [13]

Core Insights - The AI industry is experiencing significant growth, but there is a stark income disparity, with Nvidia capturing nearly 90% of market profits, leading to concerns about the sustainability of the ecosystem [3][4] - The global AI application market is projected to see substantial increases in spending, with enterprise GenAI expenditures expected to rise from $11.5 billion in 2024 to $37 billion in 2025, marking a year-on-year growth of approximately 320% [3] - The commercialization of AI applications has formed a clear hierarchy, with general large models leading the first tier, while vertical applications are rapidly gaining traction in specific sectors [5][6] Group 1: Market Dynamics - The AI application market is not as dire as perceived, with significant growth in consumer spending on applications like ChatGPT, which is expected to reach $2.48 billion in 2025, up from $487 million in 2024, representing a 408% increase [4] - The first tier of commercial applications is dominated by general large models, with OpenAI leading at an annual recurring revenue (ARR) of $10 billion and a projected compound annual growth rate (CAGR) of 260% from 2023 to 2025 [5] - Chinese applications are currently positioned in the second tier, with ARR between 100 million and 1 billion yuan, focusing on vertical applications that demonstrate clear cost reduction benefits [5][8] Group 2: Application Development - Over 200 AI applications have been launched between July and November, with a significant focus on vertical applications that address specific user needs, such as AI image processing and efficiency tools [6] - In the global top 50 generative AI apps, 22 are developed by Chinese teams, indicating that Chinese applications are competitive, although there remains a significant income gap compared to the U.S. market [8] - The cost of producing AI dynamic animations has drastically decreased, with production costs now ranging from 50,000 to 100,000 yuan, only 10% to 30% of traditional methods [17] Group 3: Challenges and Opportunities - Quality remains a major bottleneck for AI applications, with 33% of respondents identifying it as the primary challenge, particularly in terms of accuracy and consistency of output [11][13] - The current landscape shows that AI applications are primarily limited to high-cost scenarios like programming and customer service, with significant cost-saving potential but insufficient revenue generation [14] - The AI industry is moving towards a phase where understanding AI's application in business is crucial, as evidenced by the rising interest in AI-driven content creation, particularly in the animation sector [16][19]

2025 年，让 Agent 实际投产、落地应用的最大障碍已经不再是成本问题了，而是「质量」。如何让 Agent 输出可靠、准确的内容，仍然是最难的部分。 近期，LangChain 通过对工程师、产品经理、企业高管等 1300 名行业人士进行调查，深度调研了 AI Agent 目前最真实的应用情况。 进入 2026 年，企业对于 Agent 的讨论焦点，已经从「要不要做」全面转向了「如何规模化、可靠且高效地用好」。 6 个关键结论： ⬆️关注 Founder Park，最及时最干货的创业分享 超 17000 人的「AI 产品市集」社群！不错过每一款有价值的 AI 应用。 邀请从业者、开发人员和创业者，飞书扫码加群： 进群后，你有机会得到： 01 规模越大的企业， 落地 Agent 速度越快 调研数据显示，超过一半（57.3%）的受访者已经将 Agent 投入实际生产，另有 30.4% 的人正在开发且有明确的上线计划。 这一数字比去年的 51% 有了明显增长，行业正在从「概念验证」快速迈向「价值实现」阶段。 规模越大，行动越快 客户服务（26.5%）成为最普遍的 Agent 用例，研究与数据分析（24.4%）紧 ...

RockFlow 投研团队第一时间对这个数万字的系列报告进行了深度编译与逻辑重塑。我们剥离了繁杂的术语，为你精炼出决定未来十年投资格局的五大核心 叙事。Enjoy AI 基础设施与 Agent：从"交互工具"进化为"自主生命体" 划重点： 1）AI 正在从"数字助理"进化为"自主执行集群"。2026 年将见证 AI 从"对话工具"向"多智能体系统（Multi-Agent）"的跨越。a16z 预言屏幕时代 即将终结，Agent 原生基建将重定义云端速度，开启企业运营杠杆的历史性飞跃。 2）科技正在溢出屏幕，"比特"开始全面接管"原子"。电气化、材料科学与 AI 融合而成的"电子工业堆栈"将成为物理世界运行的底层逻辑。通过 软件定义制造与 AI 自动化，美国有望迎来工厂复兴的黄金时代。 3）SaaS 正经历从"被动记录"到"主动推理"的范式转移，个性化服务将实现从"为所有人优化"到"为每个人定制"的飞跃。加密货币将化身为互联 网的基础结算层，稳定币与 RWA 将重构金融底层；而预防性医疗将开启长效变现的新蓝海。 在美股市场，预见趋势的能力往往决定了 Alpha 的成色。作为硅谷风投界的"定海神针"，a16z（An ...

随着 AI 技术从"工具化"向"自主化"严谨，智能体（Agent）正在成为企业应用大模型的重要形态。那 么，如何优化 Agent，让它变得更可信、更好用，最终能够成为企业优秀的"数字员工"？ 近日 InfoQ《极客有约》X AICon 直播栏目特别邀请、RBC senior application support analyst 马可薇担 任主持人，和值得买科技 CTO 王云峰、商汤科技大装置事业群高级技术总监鲁琲、明略科技集团高级 技术总监吴昊宇一起，在AICon 全球人工智能开发与应用大会 2025 北京站即将召开之际，共同探讨如 何提升企业 Agent 的"可信度"。 部分精彩观点如下： 以下内容基于直播速记整理，经 InfoQ 删减。 定义 Agent 的技术边界 马可薇：很多人觉得 Agent 就是 Chatbot 加了几个插件。但从技术架构视角看，当系统目标从"对话"变 成"行动"，你们认为技术栈上产生的最大一个质变是什么？ 完整的过程包括：模型接收任务，判断应采取的行动，感知外界、接收反馈，并基于反馈不断调整规 划。这与过去单纯的 chatbot 模式有巨大差异，其技术复杂度和对生态的要求都远高 ...

Core Insights - The recent paper by Google DeepMind and Google Research challenges the prevailing belief in the AI community that "more agents are better" [3][5] - The research indicates that blindly increasing the number of agents is not only costly but also ineffective, with a critical conclusion that 3-4 agents represent the optimal number for current technology [3][35] Findings on Agent Systems - The "scale paradox" suggests that as task complexity increases, having more agents can lead to quicker failures, with 3-4 agents being the "golden ratio" [3][6] - There is diminishing marginal returns for agents; if a single agent achieves over 45% accuracy, adding more agents can result in negative returns [8][10] - The effectiveness of multi-agent systems is contingent on task characteristics, emphasizing the importance of matching architecture with task attributes rather than merely increasing agent numbers [4][14] Three Fundamental Laws Governing Agents - More tools lead to higher chances of "crashing" in multi-agent systems due to increased communication costs, especially when tasks require more than 16 tools [6][7] - Stronger individual agents reduce the utility of adding more agents, as communication and alignment costs outweigh benefits [8][9] - Different collaborative architectures have varying error amplification effects, with independent multi-agent models amplifying errors by 17.2 times compared to centralized models, which control errors to 4.4 times [11][12] Task and Architecture Compatibility - Multi-agent systems are not universally beneficial; their performance is highly dependent on the compatibility of architecture with task requirements [13][14] - Tasks can be categorized into three types based on their interaction with multi-agent systems: - Tasks that can be decomposed benefit from multi-agent collaboration, showing performance improvements of up to 80.9% [15] - Tasks with strict sequential dependencies suffer performance declines of 39% to 70% when using multi-agent systems [16][18] - Tasks that require both exploration and execution show mixed results, with performance varying significantly based on architecture design [19][21] Economic Analysis of Multi-Agent Systems - Multi-agent systems exhibit a drastic drop in efficiency, with centralized architectures achieving only 21.5 successful outcomes per 1000 tokens compared to 67.7 for single agents [30] - The number of communication rounds increases quadratically with the number of agents, leading to shallow reasoning and declining performance [31][34] - The optimal number of agents is identified as 3-4, beyond which communication costs dominate and lead to negative marginal returns [35]