编译：Haozhen、ChatGPT AI 正在成为企业软件史上扩散速度最快的一次技术浪潮。 过去十年，企业软件的竞争优势往往掌握在传统巨头手中，它们拥有成熟的分发渠道、深厚的数据 积累、稳固的客户关系以及强大的销售网络。但在增长最快的 AI 应用领域，竞争格局正在发生逆 转：AI-native 初创公司凭借更高的执行效率和快速迭代能力，正在实现对传统企业的超越。 这份最新的研究报告是 Menlo Ventures 基于对 495 位美国企业的 AI 决策者（包含 C-level 高管、工 程与产品副总裁等）调研的观点洞察，这些数据直观地展示了企业为何购买 AI，钱都花在了哪 里，哪些公司又从中真正受益，以及 LLM 与 AI infra 的竞争格局会如何演进等关键问题。 • 企业级 AI 的市场规模已在两年内从 17 亿美元跃升至 370 亿美元， 较去年的 115 亿美元增长约 3.2 倍， 增长速度超过历史上任何一个软件品类； • 一旦企业开始评估某项 AI 解决方案，47% 的 AI 交易最终会进入生产环境，而传统 SaaS 的这一 比例仅为 25%； • AI 应用和 infra 在 2025 年分 ...

我们相信 26 年会是多模态技术的大年，其中视频生成会快速进步让应用大规模落地，而世界模型 则会有研究上的科学突破，甚至开始从 research 走向 production。 在相当长的一段时间内， World Model 这一概念始终处于较为混沌的状态；直到近半年，随着技术 路径逐渐收敛，尤其是在具身智能与真实交互场景中出现了初步落地的案例，世界模型的轮廓开始 变得清晰。 作者：Cage、Haozhen 如果和语言模型对比：语言模型解决的是语义层面的压缩和推理，预测下一个 token；世界模型是 在解决下一步更根本的问题，AI agent 是否能真正理解时间与空间，并进行预测下一帧、下一个行 动。如果和视频生成模型对比：世界模型在交互性、实时性、长时记忆和物理合理性这四点上都需 要更进一步。 于是行业中的玩家开始在这些提升方向有了各自的 bet， World Model 领域逐步分化出两条路线： 一条以实时视频生成为核心，服务文娱、游戏等 for human 的消费者场景；另一条以显式 3D 结构 为中心，服务机器人、自动驾驶等 for AI 的领域。 本文沿着这个路线分化展开，拆解两条路线的技术趋势和落地 ...

去年此时，我们在 2025 AI Best Ideas 中提出了 20 个关键预测，12 个月过去，当我们重新审视这份 清单，惊奇地发现：我们对行业格局和技术路径的判断绝大部分都应验了，但其中也有对技术进 步、基建成熟度、以及 AI 交易复杂性的过分乐观。 2026 年即将在 "AI Bubble"的争议中开启。World Model、多模态、机器人以及新范式等积极信号的 出现意味着 AI 领域一定会出现更多惊喜，但同时，OpenAI $1.4T CapEx 也意味着，在 price-in 了 超高预期后，市场对 AI 领域的期待只会越来愈高。 以下是拾象团队 2025 预测复盘， 我们也很希望听到大家对于 AI 领域 2026 的期待， 请留下你的答 案与思考，我们会基于大家的回答组织一场"2026 AI Best Ideas"讨论。 ⬇️ 滑动或点击查看大图 ⬇️ ⬆️ 滑动或点击查看大图 ⬆️ 01 微软投资 Anthropic，模型与云格局彻底改变 ✓ 预测结果：正确 拾象预测： 2025 年 OpenAI 会变成盈利组织，微软也很有可能会投资 Anthropic，这将彻底改变模型 和云的格局——An ...

编译：Haozhen、Gemini 而支撑这场豪赌的理由之一就是 General Intuition 拥有一个业内无法复制的独特数据集。 General Intuition 是从游戏高光片段剪辑平台 Medal 中分拆而来，拥有超过 38 亿个游戏短视频片 段。与传统机器人或仿真数据不同，Pim 认为高光片段是人类在模拟环境中的情景记忆（Episodic Memory），是对人类直觉、反应和决策最密集的数字化记录。 如果说 OpenAI 通过 ChatGPT 解决了人类的"认知与逻辑"，让机器学会了像人类一样进行复杂思 考、推理与 coding，那么 General Intuition 希望赋予机器像人类一样的"直觉和物理常识"，使机器 能够在本能层面理解物理世界的空间关系。 在 CEO Pim de Witte 的构想中，LLM 负责思考与规划（Next Token）， General Intuition 则基于自 身的数据优势承担行动与交互（Next Action），两者形成互补的智能结构。团队 希望从游戏场景起 步，经由模拟环境走向自动驾驶，再延伸至机器人与物理世界，终极愿景就是实现"Atoms to ...

About Us 我们是一个对 AI 和 foundation model 痴迷的团队。 2022 年秋，我们在硅谷看到了 AI 的火苗，从此只专注研究 AI。 专注研究和投资 AI，让我们取得了还不错的成绩。我们在管 AUM 超过 15 亿美金，有 5 亿美元在投的长 线基金，一二级联动，有足够的子弹抓住 AI 机会。 我们过去投资并见证了 6 家 portfolio 从数十亿，数 百亿美金，成长为千亿美金公司——这也是拾象的寓意，只研究全球最重要的技术变化，投资有大象级 潜力的公司。 我们有数位千亿美金业务的 CEO / leadership 提供洞察，帮 portfolio 做好 AI 和全球化。 我们通过海外独角兽和 AI 讨论社群，持续讨论重要问题，帮助和影响了中美两地的华人创业者，也在 AI 从业者中获取了一些宝贵的信任。 现在，我们希望邀请你加入，一起做 全球 AI 投资，一起捕捉大机会，成为 AI 领域的最佳投手 。我们是 一个年轻（平均年龄不到 30 岁）、扁平、高人才密度的团队，推崇 high-trust，low-ego，团队内信息极 度透明，讨论氛围热烈。 我们尤其喜欢的特质是：对 AI ...

编译：Haozhen、Gemini 如今 LLM 的语言理解与生成能力已展现出惊人的广泛适用性，但随着 LLM 的发展，一个事实越 发凸显：仅靠语言，仍不足以支撑真正的智能。 从更本质的角度看，人类处理世界的方式从来不只依赖文字，而是通过视觉、空间感知、物理直觉 与行动能力等共同构成完整的认知体系。语言只是对三维世界的"有损压缩"：它记录结论，却省略 过程；它表达结构，却隐藏动态。而真正的智能，源于不断与世界互动、不断在空间中推理和行动 的能力。 正因如此，构建能够"理解并操作世界"的空间智能（Spatial Intelligence）与世界模型（World Models）成为继 LLM 之后的关键方向。 2024 年，李飞飞、Justin Johnson 等学者创立了 World Labs，今年 11 月推出了 Marble 这个 3D 世界 生成模型。团队尝试突破模型"只懂文本"的限制，让模型具备在三维环境中定位、推理、模拟、生 成甚至执行任务的能力。这不仅意味着新的技术路线，也意味着新的 AI 价值尺度：从语言走向世 界、从描述走向交互、从静态认知走向动态智能。 本文整理了李飞飞和 Justin Joh ...

Core Insights - The AGIX index aims to capture the beta and alphas of the AGI era, which is expected to be a significant technological paradigm shift over the next 20 years, similar to the impact of the internet [2] - The "AGIX PM Notes" serves as a record of thoughts on the AGI process, inspired by legendary investors like Warren Buffett and Ray Dalio, to witness and participate in this unprecedented technological revolution [2] Market Performance - AGIX recorded a weekly performance of 6.00%, a year-to-date return of 26.73%, and a return of 74.56% since 2024 [4] - In comparison, QQQ, S&P 500, and Dow Jones had year-to-date returns of 21.13%, 16.45%, and 12.16% respectively [4] Sector Performance - The application sector saw a weekly performance of 2.20% with an index weight of 33.62% - The semi & hardware sector had a weekly performance of 1.76% with an index weight of 24.22% - The infrastructure sector recorded a weekly performance of 2.08% with an index weight of 37.19% [5] AI Industry Developments - Ilya's recent interview sparked significant market discussion, highlighting concerns about model training stagnation while also noting advancements in Google's Gemini 3 capabilities [9][10] - The AI industry is transitioning from a research phase to a focus on productization and optimization, with Google leveraging its TPU technology for enhanced performance [10] - The future of AI may not be dominated by a single model but rather by productization capabilities and external factors such as distribution and ecosystem [11] Investment Trends - The AI startup financing landscape remains robust, with 49 companies securing over $100 million in single rounds by November, matching the total for 2024 [17] - Major investments include Anysphere's $2.3 billion funding round and OpenAI's record $40 billion financing, indicating a growing concentration of capital in the AI sector [17] Corporate Actions - ServiceNow is in talks to acquire cybersecurity startup Veza for over $1 billion, which would enhance its identity management capabilities [19] - Zscaler reported strong Q1 results but saw its stock drop over 7% due to a conservative outlook, reflecting investor expectations for tech company growth [19]

Group 1 - The core argument of the article highlights the structural challenges in the U.S. healthcare system that hinder drug development and commercialization, while Eli Lilly has successfully navigated these challenges through innovative GLP-1 drugs and strategic business models [2][3] - Eli Lilly's market capitalization is approaching $1 trillion, driven primarily by the success of GLP-1 drugs, which contribute approximately 80% of the company's value and have a revenue growth rate of 40% this year [4][5] - The company has a significant market share of about 70%-75% in the new patient market for GLP-1 drugs, reflecting strong investor confidence in its R&D capabilities [4] Group 2 - GLP-1 drugs help reduce daily caloric intake by approximately 800 calories, stabilizing weight loss and reducing emotional burdens associated with dieting [8] - Despite the effectiveness of GLP-1 drugs, their current usage is limited, with only about 10 million users in the U.S. compared to a potential market of 100 million obese individuals, primarily due to insurance coverage issues [11] - The direct-to-consumer (DTC) model through LillyDirect has allowed Eli Lilly to bypass traditional intermediaries, significantly increasing efficiency and revenue, with annual income reaching billions [24][26] Group 3 - Eli Lilly's R&D spending is projected to reach 20%-25% of sales, approximately $14 billion, which is comparable to national research institutions [18] - The average cost of developing a new drug is estimated at $3.5 billion to $4 billion, with over 60% of this cost attributed to late-stage clinical trials [19] - The company employs a mixed model of internal R&D and external collaborations to balance innovation and efficiency [22] Group 4 - The U.S. healthcare system faces significant structural issues, including a misalignment of funding for chronic disease management and a reliance on high-cost acute care [28][29] - The generics market, while providing low-cost medications, suffers from quality inconsistencies and supply risks, which can affect patient outcomes [30][31] - Regulatory requirements for new drug approvals have become increasingly stringent, extending development timelines and costs significantly [35][36] Group 5 - The pricing of drugs in the U.S. is often opaque, with significant discrepancies between list prices and actual transaction prices, leading to challenges for smaller institutions in negotiating fair prices [46][48] - Traditional pricing models do not adequately address the value of innovative therapies, such as gene therapies, which require new pricing strategies to reflect their long-term benefits [49][50] - Clinical trial costs are rising, with median costs per participant exceeding $40,000, driven by the complexity of patient recruitment and the need for high-quality care [53][54]

Core Insights - Gemini 3 represents Google's significant return to leadership in the AI space, marking the beginning of a new competitive landscape among major players like OpenAI and Anthropic [4][14]. Group 1: Model Strength and Capabilities - Gemini 3's training FLOPs reached 6 × 10^25, indicating a substantial investment in pre-training compute power, allowing Google to catch up with OpenAI [5][6]. - The model's data volume is speculated to have doubled compared to Gemini 2.5, providing a significant advantage in pre-training and creating a strong intellectual barrier [7]. - Gemini 3 employs a Sparse Mixture-of-Experts (MoE) architecture, achieving over 50% sparsity, which allows for efficient computation while maintaining a vast parameter space [10][11]. Group 2: Competitive Landscape - The competitive landscape is evolving into a dynamic structure where Google, Anthropic, and OpenAI alternate in leadership positions, reflecting their differing technological and commercial strategies [14][15]. - Google has a cost advantage in inference due to its proprietary TPU cluster, while its coding capabilities are on par with OpenAI and Anthropic [15][17]. Group 3: Benchmark Performance - Gemini 3 outperformed its competitors in various benchmarks, achieving 91.9% in scientific knowledge tests and 95.0% in mathematics without tools, showcasing its superior reasoning capabilities [16]. - In terms of speed, Gemini 3 processes tasks approximately three times faster than GPT-5.1, completing complex tasks at a significantly lower cost [22]. Group 4: Organizational and Developmental Insights - The successful integration of DeepMind and Google Brain has led to improved model iteration speeds, overcoming previous internal challenges [13]. - Google has developed a unique "product manager-style programming" approach, enhancing user interaction and project management during coding tasks [12]. Group 5: Commercialization and User Engagement - Google is prioritizing user experience over immediate monetization, focusing on long-term user retention and ecosystem health [61][68]. - The introduction of tools like Antigravity and the integration of Gemini into Chrome are strategies to enhance user engagement and capture valuable feedback for model improvement [62][64]. Group 6: Future Prospects and Market Dynamics - The shift towards multi-modal capabilities in AI, as demonstrated by Gemini 3, positions Google favorably in the evolving landscape of AI applications, particularly in video generation [25][45]. - Google's TPU technology is projected to significantly reduce model training and inference costs, potentially disrupting Nvidia's dominance in the market [46][49].

Core Insights - The AGIX index aims to capture the beta and alphas of the AGI era, which is expected to be a significant technological paradigm shift over the next 20 years, similar to the impact of the internet [2] - The "AGIX PM Notes" serves as a record of thoughts on the AGI process, inspired by legendary investors like Warren Buffett and Ray Dalio, to witness and participate in this unprecedented technological revolution [2] Market Performance - AGIX experienced a weekly decline of 5.65%, with a year-to-date return of 19.56% and a return of 64.68% since 2024 [4] - In comparison, QQQ, S&P 500, and Dow Jones also saw declines of 3.09%, 1.95%, and 1.91% respectively during the same period [4] Sector Performance - The Application sector declined by 1.36%, the Semi & Hardware sector by 1.14%, and the Infrastructure sector by 2.50% [5] AI Industry Developments - Microsoft announced a comprehensive autonomous security solution to address security challenges posed by the large-scale deployment of AI agents, enhancing enterprise network defenses [16] - Alphabet's Intrinsic and Foxconn formed a joint venture to develop next-generation intelligent robotic systems, combining AI-driven software with smart manufacturing platforms [17] - Amazon's Prime Video introduced an AI-generated "video recap" feature, showcasing advancements in AI applications within the film industry [18] - Cloudian launched the HyperScale AI data platform, designed to convert unstructured data into AI insights, addressing challenges faced by enterprises in adopting AI [18] - Adobe announced the acquisition of Semrush for approximately $1.9 billion to strengthen its marketing product offerings in response to the AI search transformation [18] - Cloudflare acquired AI deployment platform Replicate to enhance its AI inference service capabilities [19]