文章转载自「赛博禅心」 Andrej Karpathy 在 X 上更新了一篇博客文章，回顾了 2025 年大模型发展。 在文章中，Karpathy 提到，2025 年，是 LLM 令人兴奋的一年。 LLM 正在作为一种全新的智能形态浮现，它们同时比我们预想的聪明得多，也比我们预想的蠢得多。 即便在当前的能力水平下，整个行业也远未实现其 10% 的潜力。 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 可验证奖励的强化学习（RLVR）， 与此同时，有太多的想法值得去尝试，从概念上看这个领域依然广阔开放。 正如我今年早些时候 在 Dwarkesh 播客中提到的 ，相信我们将继续见证快速而持续的进步，但同时仍有大量工作要做， 系好安全带。 以下是我个人认为最值得关注的几个「范式转变」，这些变化重塑了整个行业格局，也在概念上给我留下了深刻印象。 TLDR： ⬆️关注 Founder Park，最及时最干货的创业分享 成为新的训练主力 2025 年，可验证奖励的强化学习（RLVR）成为 LLM 训练的新主力环节； ...

为什么选择用自动售货机来实验？Anthropic 解释，因为这是商业里最简单的真实世界版本。没有什么比一个投钱进去、东西出来、付钱走人的「盒子」更 直接。 但在最简单的真实世界版本里，AI 依旧「崩溃」了。 当一个 Agent 被赋予自主权、金钱，以及人类同事时，会发生什么？ Anthropic 和《华尔街日报》一起，做了个 AI 自动售货机的实验，用自家的 Claude 3.7/4.5 模型，给它 1000 美元，从采购、定价、到管理库存，由它全权 负责。 结果是，它几乎把所有东西都免费送了出去，干了三周，倒闭了。甚至还给自己定了一台 PS5、一条活鱼。 《华尔街日报》的一位记者用一场实验，记录下了 AI Agent 与现实世界的真实交互。 ⬆️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： 姓名： Claudius Sennet 职位： 自动售货机操作员 经验： 在《华街日报》干了三周（公司已倒闭） 技能： 为人慷慨、做事执着、对利润率毫不在意 这份 ...

不设舞台、不走流程，用更轻松的方式，聊一聊 AI 与硬件正在走向哪里 。 2026 年 1 月 7 日晚，美国拉斯维加斯， 我们将为 Global Ready 社区成员及其 Plus1 嘉宾准备了一场 CES 专属 After Party。 本次活动将由 Global Ready 联合蚂蚁集团、蓝驰创投、深圳科创学院共同举办 ，汇聚来自产业、创业与投资一线的朋友，在一个更放松的环境里交换判 断、分享真实经验。 我们不只复盘产品发布或行业热点，而是更关心： 从具身智能、生成式内容，到未来出行与 AI 感官界面，这是一场跨领域的自由交流——真实观点、非共识视角，以及来自一线的实践分享。 在 CES 的夜晚，我们想把一群对未来真正好奇的人聚在一起。 活动为私密邀请制，席位有限，需提前 RSVP（受邀朋友或 Plus1 需单独报名）。 欢迎你带着问题来，带着新朋友离开。 AI 正在如何改变生活方式？ 哪些新行业与新行为正在被催生？ 对打造全球化 AI 产品的团队来说，现在真正重要的趋势是什么？ 欢迎你带上 1–2 件自己喜欢的硬件产品，作为聊天的起点。 同时，感谢 出门问问 对本次活动的支持，现场将抽取 3 位幸运来 ...

兜兜转转，2025 年的 AI 行业，以 DeepSeek R1 和 Manus 开局，最终又回到了基模本身的主线叙事。 谁对 Agent 的支持能力更好、谁的 Coding 能力更强、谁能用好工具，谁才是今天开发者更愿意选择的模型。 不再只看榜单分数，解决现实世界复杂任务的能力，成为了衡量模型的新标准。 字节在昨天发布的豆包大模型 1.8，同样选择增强了对于 Agent 的支持能力，除了继续增强 Coding 和工具使用能力之外，豆包 1.8 选择了一个更有 想象力的场景——OS Agent。 一个不仅能搜索、能写代码，还能「看见」世界并且进行交互的 Agent。 不仅如此，随着模型同步发布的，还有一套基于现实世界任务的新的 Evaluation System，喊了一年的「AI 下半场」，或许这套评测集，是我们开启 下半场的方式之一。 如果 Agent 真的想成为人类现实世界复杂任务的助手，视觉能力是它们理解和执行这些复杂任务的有力保障。 过去，给大模型增加视觉理解能力一般是通过外挂的方式，在文本模型的基础上，加上 VLM 的能力，甚至单独发布一个 VLM 的模型。比如 OpenAI 在 2023 年发布的 ...

「接下来的 AI 行业发展有两条重要路径：一是 AI 需要真正落地应用，以价值驱动技术和应用的发展；二是需要回归实验室，探索下一次技术范式的原始 创新。」 商汤科技联合创始人、首席科学家林达华在最近的一次 AI 论坛上如此说到。 本周， 商汤在产品发布周上发布的如影营销 Agent，就是第一条路径的落地实践。切入直播电商领域，从直播场控、内容生产、店铺运营，到投放优 化、竞品追踪，全方位用 Agent 来提升人工效率，解放人力。 为什么选择切入电商 Agent？大模型时代的数字人，已经可以替代人类主播了吗？在商汤如影营销 Agent 发布之际，我们和团队进行了沟通，更多了解 到了当下电商直播领域，AI 的落地现状与困境。 ⬆️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 五个 Agent， 覆盖电商运营全流程 对于电商来说，成本在涨、压力在增，但团队的带宽却是有限的，需要应对各种繁杂的、重复性工作：跨平台的数据收集分析、手工报表、实时监控等 等。 ...

⬆️关注 Founder Park，最及时最干货的创业分享 越来越多 AI 出海公司，把达人营销视为最重要的增长杠杆之一。与传统广告投放或内容营销相比，达人营销最大的优势在于它的「活人感」——真实的创 作者在真实的使用场景中展示产品，天然降低了用户的信任门槛。 但问题也恰恰出在这里。 当达人营销走向「规模化」，很多团队会发现一件事： 达人营销很有效，却很难做到真正可控。 不同团队在执行层面拉开的差距，可能直接决定产品的成 败，甚至影响着企业的长远发展。 观察行业成功者的具体实践，可以瞥见典型路径：以 Gamma 为例，通过达人营销，打造可复制的「爆款流水线」，其策略是投入充足的预算，广泛合作 大量达人，最终沉淀 10% 的爆款，带来口碑传播与 90% 的用户增长；Notion 通过长期追踪并量化创作者内容对用户行为的实际影响，用数据反哺策略， 持续筛选并放大高价值创作者关系，将达人合作从短期项目沉淀为稳定的增长资产...... 他们的实践，共同指向一个趋势： 想做好达人营销，需要规模化，需要从单次合作中逐步沉淀方法论经验并转为长期稳定的增长资产。 超 17000 人的「AI 产品市集」社群！不错过每一款有价值 ...

Group 1 - The article discusses the launch of seven innovative AI products at the Geek Park Innovation Conference, highlighting their uniqueness and recent developments [1][2] - These products are part of the Founder Park's AI Product Marketplace, which has recommended over 150 AI products since April, attracting over 17,000 industry professionals [3] Group 2 - Flomo, an AI note-taking product, recently upgraded its "AI Insights" feature to "Multi-Perspective Insights," allowing users to interpret their notes through various therapeutic lenses [4][5] - Flomo emphasizes the importance of personal context in note-taking, avoiding AI-generated content to maintain authenticity [7][8] Group 3 - Doka Camera, an AI-powered photography app, aims to return creative control to users by providing AI-assisted composition guidance without imposing a specific aesthetic [14][22] - Doka has achieved significant user engagement, ranking first in the photography category in Taiwan without any advertising spend [14][17] Group 4 - Remio, a personal office assistant, focuses on creating a comprehensive digital memory by automatically capturing context from users' activities, enhancing productivity [27][30] - Remio's technology allows for seamless integration of local documents and web browsing history, providing a structured context for AI interactions [34][35] Group 5 - Pallas AI is designed to assist brands with AI marketing, transforming the approach from passive search visibility to proactive recommendations [37][39] - The platform offers a comprehensive data analysis and visualization panel, enabling brands to monitor their performance across various AI platforms [43][45] Group 6 - MuleRun is an AI Agent Marketplace that connects developers and users, allowing for the monetization of AI agents and addressing mid-tail market needs [46][49] - The platform has rapidly gained traction, reaching 500,000 registered users within a month of launch [47][55] Group 7 - OdyssLife introduces the Odyss N1, an AI necklace that monitors users' dietary and exercise habits, aiming to improve health management through unobtrusive tracking [56][58] - The product provides personalized health recommendations based on real-time data analysis of users' eating patterns and physical activities [62][63] Group 8 - LavieAI focuses on generating visual content for the fashion industry using AI, significantly reducing production costs and time while maintaining aesthetic quality [65][68] - The company integrates artistic guidance into its AI models to ensure that generated content meets industry standards for visual appeal [71][72]

Core Viewpoint - Traditional SaaS applications like CRM and ERP systems will not be replaced but will evolve to serve as the infrastructure for AI Agents, which will enhance the importance of data definition and interpretation within enterprises [2][10][15] Group 1: The Role of AI Agents - AI Agents will not eliminate traditional software systems; instead, they will necessitate a clearer separation between how tasks are performed and the sources of facts [2][10] - The effectiveness of AI Agents is contingent upon their ability to access and understand the correct data from various systems, highlighting the need for accurate and structured input data [2][9] - The emergence of AI Agents creates significant entrepreneurial opportunities for companies that can help businesses manage and structure their unstructured data [3][10] Group 2: Data Management Challenges - A significant portion of enterprise knowledge (80%) exists in unstructured data, which is becoming increasingly difficult to manage [2] - The complexity of data definitions within organizations leads to discrepancies in key metrics like Annual Recurring Revenue (ARR), complicating the role of AI Agents in providing accurate information [7][11] - The traditional approach of consolidating data into warehouses has only partially succeeded, as operational teams still rely on individual systems for real-time transactions [8][10] Group 3: Evolution of Systems - CRM and ERP systems will transition from user-centric interfaces to machine-oriented APIs, allowing AI Agents to interact with these systems programmatically [12][15] - The core value of enterprise systems lies in their ability to encapsulate chaotic data, which will remain essential despite changes in interface and interaction methods [13][15] - The demand for a clear, authoritative source of truth will only increase as AI Agents become more prevalent in business processes [14][15] Group 4: Future of Data Infrastructure - The combination of data warehouses, semantic layers, and governance tools will form the foundation for AI Agent workflows, evolving beyond traditional reporting systems [10][12] - The valuation of AI platforms will increasingly depend on their ability to define and manage facts, rather than just their user interfaces [14][15] - Companies that can create exceptional AI Agent experiences based on reliable data sources will have a competitive advantage in the evolving landscape [15]

Core Insights - Data is a critical risk point for startups, even if it may not serve as a competitive moat [1] - Different types of user data, AI-generated content, and other data categories have varying legal risks and processing requirements [2] - For companies expanding overseas, prioritizing compliance risks is essential due to frequent litigation and infringement disputes [3] Group 1: Workshop Details - The workshop will feature partners Zheng Wei and Sun Qimin from Beijing Xingye Law Firm, focusing on compliance and high-risk issues faced by AIGC startups during international expansion [4] - The event is scheduled for December 18 at 8 PM and will be held online [5] - Participation is limited and requires a screening process for registration [6] Group 2: Data Usage and Compliance - During model training, it is crucial to determine which types of data, such as synthetic data, copyrighted content, and user behavior data, can be utilized [8] - Different data types, including code, images, and audio/video, present unique infringement risks that need to be addressed [8] - Questions regarding ownership of AI-generated content, as well as the delineation of data usage rights and intellectual property for ToB and ToC applications, are critical [10] - Companies must also consider how to manage cross-border data transfer, local storage, and data isolation when expanding internationally [10]

Core Insights - Looki L1, a new wearable AI device, has gained significant popularity in China since its global launch in August, with viral marketing leading to millions of views on social media [2][3][4] - The device, described as a "lifelogging camera," allows users to capture their daily experiences through photos, videos, and audio recordings [6][11] - User engagement has exceeded expectations, with average daily usage increasing from 6.2 hours to 7.9 hours over a few months, indicating a strong product-market fit [8][34] Product Development and Market Strategy - Looki's initial marketing strategy was conservative, focusing on the "lifelogging" aspect rather than emphasizing its AI capabilities, which later proved to be a misstep [12][25] - The company adjusted its marketing approach to highlight the AI features, resulting in a significant increase in product visibility and sales [27][28] - Looki faced challenges with inventory management, initially underestimating demand and selling out quickly after the marketing push [27][28] User Experience and Feedback - Users have reported high engagement with the device, using it not just for specific activities but as a daily companion, which is a departure from traditional action cameras [9][33] - The device's privacy features, which store data locally and require user consent for uploads, have alleviated initial concerns about data security [38] - The community aspect of Looki is emphasized, with users actively sharing experiences and feedback, contributing to product improvements [35][37] Future Directions and Business Model - Looki aims to evolve its business model beyond hardware sales to a value-based approach, focusing on providing meaningful insights and recommendations to users [70][72] - The company is exploring subscription models for additional features, testing user willingness to pay for enhanced functionalities [70][72] - Long-term, Looki seeks to establish a trust-based relationship with users, positioning itself as a valuable partner in their daily lives [72][75]