「Jinqiu Update」 更新锦秋与被投公司12月-1月新鲜进展 < 锦秋基金出品的视频播客栏目 > 「锦供参考」 上线 Vol.02 ｜ 180 万个动画超级个体的诞生 shownotes 当执行的门槛被技术削平，动画是否能成为每个人的"超级表达"？ 本期「锦供参考」，我们邀请到了 OiiOii 创始人及"小学校长"闹闹 。作为视觉产品领域的顶尖幕后推动者，闹闹曾先后在 微信、字节跳动、阶跃星辰 深度主导多 款国民级视觉产品的研发。她曾是 剪映和抖音特效 背后的核心"魔法师"，亲手定义了数亿人的创作方式。 在闹闹看来，OiiOii 并非单纯的生成工具，而是一套由 AI 编剧、分镜、音效组成的 智能协作 Agent 系统， 正在让 180 万个"动画超级个体"成为可能。锦秋基金 创始合伙人杨洁、合伙人臧天宇将与闹闹深度对谈，拆解这场关于"身份翻盘"与 AI 生产力重构的技术实验。 Vol.01 ｜中美 AI 创业激荡 2026 ：硅谷视角、融资真相与被忽视的机会 shownotes 当 2026 年成为中美 AI 创业更加激荡的一年，中国创业者该如何在这个全球大舞台上精准定位、高速迭代？ 本期「锦供参考 ...

Core Insights - The AI funding landscape has seen a dramatic increase, with global AI financing reaching $225.8 billion in 2025, a 97% year-over-year increase, despite a slight decrease in the number of deals [3][10] - The funding is increasingly concentrated, with 78% of the capital flowing to less than 5% of projects, indicating a shift towards larger investments in fewer, high-potential teams [3][4] - The valuation dynamics have shifted from "scale-driven" to "density-driven," where smaller, more efficient teams are achieving higher valuations per employee [5][75] Investment Trends - Global AI financing reached $225.8 billion in 2025, a 97% increase from the previous year, while the number of deals slightly decreased, indicating larger average deal sizes [10][26] - In Q4 2025, AI financing surged to $83.2 billion, doubling compared to the average of the first three quarters [13] - The median deal size in the AI sector rose to $4.1 million, with the U.S. leading at a median of $5 million per deal [28][26] Regional Insights - Over 85% of global AI funding in Q4 2025 was directed towards the U.S., with a significant valuation gap between U.S. and Asian projects [4][16] - The median deal size in the U.S. was $5 million, compared to $3 million in Asia, highlighting the regional valuation disparity [4][28] Valuation Dynamics - The capital's focus has shifted to "density-driven" valuations, where smaller teams are achieving higher valuations per employee, such as OpenEvidence with a valuation of $173.9 million per employee [5][75] - In 2025, 67% of M&A targets had fewer than 50 employees, indicating a trend towards smaller, more agile companies being favored in acquisitions [68][68] Exit Trends - In 2025, there were 830 global AI exit transactions, with 782 through M&A, marking a 61% increase year-over-year and the highest in five years [94][94] - The U.S. accounted for 47% of exit transactions, demonstrating its dominance in both investment and exit activities [98][98] Talent Trends - The fastest-growing segments in AI talent include AI agent browser infrastructure (87.2% growth), voice AI development platforms (54.1% growth), and multi-agent systems (49.8% growth) [66][66] - Talent is migrating from foundational model layers to application infrastructure, indicating a shift in focus within the AI industry [7][7] Unicorn Insights - In 2025, 75 new AI unicorns were created, with the U.S. dominating the landscape, holding 256 out of 360 total unicorns by year-end [80][81] - The top three newly minted unicorns in Q4 2025 were Mercor ($10 billion), Metropolis ($5 billion), and Unconventional AI ($4.5 billion) [90][90]

「Jinqiu Select」 < Overview > 在 AI 时代，这意味着虽然单行代码的编写速度加快、成本大幅下降，但全球对软件的需求却完全没有放缓的迹象。未来的软件工程将经历本质重构， 更高比例的全球 GDP 将通过这种方式创造，且未来的软件将不再是通用型的"大厂产品"，而可能是仅为个人或极小群体定制的、持续进化的"私人软 件"。 跨越语言与时差，传递科技圈最值得被听到的声音。 更高比例的全球 GDP 将以此种方式被创造与消费。 代码免费、智力通缩，2026年后的AI创业者靠什么建立护城河？ 我们正处在一个"智力"作为生产要素被彻底重构的奇点 。 最近，OpenAI 首席执行官萨姆·阿尔特曼（Sam Altman）在内部交流会中的发言，为这种焦虑提供了一份极具清醒感的"内参" 。 Sam Altman的回答基于一个深刻的经济学逻辑——杰文斯悖论： 当技术的进步提高了某种资源的利用效率，该资源的消耗量反而会因为需求的指数级 增长而增加。 OpenAI 内部预测， 到 2027 年底，提供 GPT-5.2x 级别智力的成本将比现在降低 100 倍以上 。 这意味着"智力"将真正成为一种像电力一样廉价、按 ...

以下文章来源于脑洞航海家 ，作者Bo & Rita 脑洞航海家 . 关心 XR、AI、文娱游戏、消费科技，一起探索「Smart, Rich, & Fun」的脑洞漩涡。FA 业务可联系：melyanor 「Jinqiu Spotlight」 追踪锦秋基金与被投企业的每一个光点与动态， 为创业者传递一线行业风向。 "当执行成本归零，意图就是新的源代码；最终胜出的公司，将是那些能最好地捕捉并普及组织知识的公司。" 这是 Isoform 创始人 Bo 对未来软件开发的判断。 Bo 正在打造新一代软件协作平台 Yansu（严肃）——一个能够主动捕捉团队意图、沉淀组织知识，并将其转化 为可执行软件成果的平台。作为 MLOps 和 分布式 系统领域的资深工程师，Bo 曾联合创立了知名 大模型 应用 开源平台 BentoML， 也完成了千万美元级融资。在看到意图识别赋能ai coding可带来的巨大价值后，Bo 再度 出发，创立了 Isoform。 锦秋基金于 2024 年底完成了对 Isoform 的投资。 当时，锦秋在美国组织了一场"小饭桌"，在与现场三家公司 的创始人深度交流后，锦秋迅速决定对这三家公司同时进行投资，I ...

匿名潜伏在 OiiOii 的用户群里，闹闹看到了不少和她当年一样的人，热爱ACG，等一个合适的机会展示自己的热情与天赋：一个性格内敛的北漂女孩 用动画拍出了不敢出镜的 Vlog；一位小学老师因为做出了让学生惊叹的课件而找回了成就感；锦秋基金创始合伙人杨洁自己，也用 OiiOii 为转学的女 儿重构了一个"太空学院"的奇遇。 在闹闹看来，这不仅是工具的迭代，更是产业逻辑的重组。过去，动画是集体的重资产消耗；而未来，当内容生产的边际成本趋近于零，行业将进 入"意图主导"的 时代。 在北京那些被数据和业务填满的深夜里，闹闹偶尔会被认为是一个"潜伏者"。 过 去的十年里，她的名字出现在腾讯微信、字节跳动、B站、阶跃星辰这些互联网巨头的名人名单里。 作为一名资深产品经理，她不仅擅长用理性的逻 辑去拆解用户增长，也在长期的产品实战中，建立起一种对用户需求极其敏锐的职业直觉。 但在这些职业标签之 下，其实还一直藏着一个关于动画的人生伏笔。 十年前，闹闹曾试图入行。她辞职去学动画，却发现那是一座由复杂的软件操作、昂贵的渲染成本和冗长的工业管线筑起的围墙。那时，一个缺乏资源 支撑的个体创作者，很难获得表达的机会。此后，即便身处 ...

「Jinqiu Spotlight」 追踪锦秋基金与被投企业的每一个光点与动态， 为创业者传递一线行业风向。 锦秋基金已完成对宇树科技投资。 锦秋基金，作为 12 年期的 AI Fund，始终以长期主义为核心投资理念，积极寻找那些具 有突破性技术和创新商业模式的通用人工智能初创企业。 根据市场推算，位居行业出货量第一。 2026年1月22日 UNITREE 作为全球出货量第一的具身机器人品牌，宇树科技致力于"创建世界的科技树，用科技推 动世界进步"，专注于消费级与行业级高性能通用足式/人形机器人及灵巧机械臂的自主研发、生产和销售。公 司打造最具性价比优势的四足与人形机器人本体，提供完整的开发套件和基础运动控制能力，是全球首家公开 零售高性能四足机器人并最早实现行业落地的公司。 近日，宇树公布了年度销售数据， 2025年人形机器人出 货量超5500台。 2. 宇树2025全年人形机器人实际出货量超 5500台（指实际出售发货给终端客户的 数量,并非订单数量,订单数量更高）, 2025年本体量产下线超6500台。 3. 上述均为我司纯人形机器人的数量,不含 我司双臂轮式等其他机器人产品。 4. 目前各种机器人形态 ...

在上周五的锦秋小饭桌，我们将一个问题抛给了现场的 AI Builder 们： 如果只能给出一个判断，你会如何预测 2026？ 数小时的交流中，不同赛道的创业者给出了各自的判断。一些共识反复出现，跨越具体技术路线，指向正在发生的 结构性变化。大家从供给侧谈起，延伸到内容生产、产品护城河、定价方式、交互入口，以及在过剩环境中重新变 得稀缺的信任。 基于这次小饭桌的讨论，我们将一线 AI Builder 的观点整理为一组关于 2026 年的预言，分享给大家。 滑动查看更多现场图片 #预言1 2026 ，"AI 即供给" 在 2023 年到 2026 年这短短三年间，人工智能对全球生产供给侧的重构，并不是一夜之间发生的，而是一个从"工具"变成本质的过程。回看这段历 程，实际上是 AI 从替代局部到重构全局的演进。 第一阶段（2023-2024）是"工具化阶段"，核心是辅助与降本。 这一时期的逻辑核心在于"减人"，AI 作为一种高效插件，以 Co-pilot（副驾驶）的 身份显著提升了原有供应链的斜率，实现了生产成本的初步下降。 随后进入了 2025 年的"商品化阶段"，核心特征是分化与平替。 随着模型价格的"崩塌"， ...

「Jinqiu Spotlight」 追踪锦秋基金与被投企业的每一个光点与动态，为创业者传递一线行业风向。 计算机界的"名人堂"再次迎来闪耀时刻。 今天，全球最具影响力计算学会ACM官宣： 2025 ACM Fellow名单，正式出炉！ 清华大学计算机系教授、生数科技首席科学家 朱军当选。 作为机器学习领域的深耕者，朱军教授 在贝叶斯方法与表示学习等底层理论上贡献卓著， 凭借"稀疏主题编码"打 破传统表示学习的边界。 同时， 朱军教授也正在带领生数科技团队在多模态生成的"技术深水区"持续打磨。 锦秋基金是 生数科技天使+轮独家投资方。 依托于 2022 年全球首创的U-ViT 架构，生数科技率先发布了国内首 个视频生成大模型 Vidu。随着 2025 年视频生成技术全面渗入内容生产链路，进入 2026 年，行业的目光将从单 纯的'模型突破'转向更深层次的"场景扎根"。 在首期锦秋会@2025 Experience With AI大会上，朱军教授谈到 在进入视频生成领域时，格外敏感于"涌现"的价 值，并分享 多模态模型能力的背后，正在迎来一个"生产力跃迁"时代。 以下为新智元对 朱军、陈宝权、贾佳亚、熊辉等19 ...

Core Insights - The article discusses the introduction of the Contrastive Latent Action Pretraining (CLAP) framework, which aims to address the data scarcity issue in robot learning by leveraging abundant human behavior videos from platforms like YouTube and Douyin [4][10]. Group 1: CLAP Framework Overview - The CLAP framework aligns the motion space extracted from videos with the action space of robots, effectively avoiding the "visual entanglement" problem commonly faced by existing latent action models [9][11]. - It utilizes a unified Visual-Language-Action (VLA) framework that combines the precision of machine data with the semantic diversity of large-scale unannotated human video demonstrations [14]. Group 2: Training Methodology - The research team developed two VLA modeling paradigms: CLAP-NTP, a self-regressive model excelling in instruction following and object generalization, and CLAP-RF, a strategy based on Rectified Flow aimed at high-frequency, fine-grained control [10][16]. - A knowledge matching (KM) regularization strategy is introduced to mitigate catastrophic forgetting during the fine-tuning process, ensuring that robots retain previously learned skills while acquiring new ones [11][16]. Group 3: Experimental Results - Extensive experiments demonstrate that CLAP significantly outperforms strong baseline methods, enabling effective skill transfer from human videos to robot execution [18]. - Performance comparisons in real-world tasks show that CLAP-NTP and CLAP-RF achieve success rates of 90% and 85% respectively in pick-and-place tasks, indicating superior capabilities [20]. - Robustness evaluations reveal that CLAP-RF maintains a mean success rate of 66.7% under environmental perturbations, showcasing its resilience [21].

Core Insights - The article discusses the innovative approach of "Zao Meng Ci Yuan" (Dream Dimension), an AI content community that has successfully attracted over 10 million users, primarily young individuals born in the 2000s and 2010s, with an impressive average daily engagement time of 1 hour and 50 minutes, indicating high user retention and engagement levels [6][18]. Group 1: Company Overview - "Zao Meng Ci Yuan" is positioned as an AI content community that transforms model capabilities into consumable content experiences, focusing on user engagement and interaction [9][63]. - The company has achieved a peak daily token consumption of 20 billion, showcasing its significant demand for AI model usage [9]. - The founder, Shen Qiaojin, emphasizes the importance of character models in AI content, which are highly consumable and have led to a diverse user base [14][20]. Group 2: User Demographics and Engagement - The primary user demographic consists of young females aged 12 to 35, with a core focus on those aged 14 to 20, reflecting a trend towards younger audiences with rich imaginations [27][44]. - The platform has successfully attracted users from lower-tier cities, leveraging their imaginative capabilities while avoiding inappropriate content [44][46]. - User engagement is further enhanced by the platform's ability to adapt to various content forms, including text, music, and video, which broadens its appeal [60][62]. Group 3: Business Strategy and Market Positioning - The company aims to create a large user base by collaborating with model manufacturers to share early computational costs, allowing for sustainable growth [16][18]. - Shen Qiaojin's strategy involves continuously matching user demographics with evolving model capabilities, ensuring that the platform remains relevant and engaging [45][61]. - The focus is on creating a versatile platform that can support various content types, thus attracting a wider audience and enhancing user experience [78][80]. Group 4: Future Outlook and Innovations - The company anticipates that advancements in multi-modal capabilities will attract a broader user base, particularly those who prefer video content [60][61]. - Shen Qiaojin predicts that improvements in model capabilities will lead to richer content experiences, further expanding the user demographic [62][64]. - The long-term vision is to establish "Zao Meng Ci Yuan" as an AI-native character brand, emphasizing the importance of character IP in driving user engagement and content creation [66][70].