嘉宾：创世伙伴创投创始主管合伙人 周炜 当下，AI与机器人赛道正经历从底层技术突破到应用落地的关键转折期。作为创世伙伴创投的创始主 管合伙人，周炜以其技术出身的背景和跨越互联网、硬科技多轮周期的投资经验，对行业趋势有着独到 判断。 近日周炜与新浪科技《科创100人》栏目对话。在对话中，周炜表示，大模型行业正经历残酷洗牌，但 惨烈程度远不及当年的电商大战。未来可能仅存1-3家真正的大模型巨头，而无法跻身头部的企业将转 向垂直领域发展。周炜还谈到，真正的技术价值在于解决实际问题，呼吁投资者更关注AI应用层而非 盲目追逐底层技术热潮。 文 | 新浪科技 徐苑蕾 期数：No.74 谈及当下风险投资机构的退出机制，周炜指出，并购仍是最主要的退出方式，而IPO则象征企业进 入"成年阶段"，健康的二级市场对于投资机构的退出同样至关重要。周炜坦言，近年来，中国企业上市 数量减少，导致当前资本市场面临"堰塞湖"困境，亟需疏通退出通道以维持创投生态的良性循环。 谈大模型竞争：行业分化是必然趋势 近两年，大模型行业经历了一轮显著的洗牌，许多公司因商业化困难、资金链断裂或竞争压力倒闭或寻 求收购。行业呈现"强者愈强"的马太效应，随着 ...

作者：郑峻 外界期待看到苹果展示AI部署蓝图，但却只看到了他们更换华丽的界面外衣。 华丽外衣重磅发布 今年的WWDC是产品发布的大年还是小年？在设计语言上，这是苹果过去十多年来最重大的一次产品 更新。但在产品功能方面，今年是绝对的小年，缺乏真正的前沿科技创新。AI驱动的iPhone？9月份发 布的iPhone 17系列或许机身更轻薄，或许拍照更惊艳，但不会有什么AI功能方面的惊喜。 昨天的重磅新品无疑是全新的"Liquid Glass"设计语言，采用了更丰富的透明与流体效果，应用于电 脑、平板、手机、手表等全平台软件，带来了更为炫酷与一致等视觉体验。这是苹果自2013年从拟物化 改成扁平化设计风格之后最大的一次界面风格更新。 这一设计语言令人不禁回想起18年前的Windows Vista（没有嘲讽的意味）。虽然微软的这款操作系统 并不算成功，但其Aero Glass界面设计语言和苹果最新的Liquid Glass有着诸多相似之处，也在社交网络 上引发了诸多热议。看起来，软件界面设计风格也和时尚一样存在着潮流轮回。 没有AI可以宣布 有惊喜吗？没有。这是外界所期待的WWDC吗？科技行业已经进 入AI时代三年了， ...

Core Insights - The KubeCon + CloudNativeCon China 2025 event highlights the importance of open-source software in driving innovation within technology companies, as emphasized by Jim Zemlin, Executive Director of the Linux Foundation [1] - China is recognized for its significant contributions to cloud-native technology, being one of the earliest and strongest ecosystems in the CNCF, with notable open-source projects like Volcano and Dragonfly [1] - The transition towards AI cluster clouds necessitates major changes in cloud management and construction, as stated by Ren Xudong, Huawei's Chief Open Source Liaison Officer [2] Group 1 - Jim Zemlin encourages technology companies to adopt open-source software, citing a study indicating that the cost of acquiring all necessary open-source software would reach $9 trillion [1] - CNCF CTO Chris Aniszczyk acknowledges China's role in cloud-native innovation, noting its position as the second-largest contributor to open-source globally [1] - The event showcases various impactful projects from China, including Volcano, Dragonfly, KubeEdge, and OpenYurt, demonstrating excellence in edge computing and distributed processing [1] Group 2 - Ren Xudong discusses the evolution of AI hardware infrastructure towards large processor clusters, emphasizing the need for significant changes in cloud management [2] - Huawei's software engineer Xuzheng Chang reports that iFlytek improved GPU utilization by over 40% and reduced fault recovery time by 70% using Volcano for large-scale model training [2]

Core Viewpoint - Didi Enterprise Edition launched an "AI + one-stop" business travel solution to address industry pain points such as fragmented travel processes, poor experiences, and uncontrolled costs [1][2] Group 1: Industry Context - The business travel industry is entering a "precise travel" era driven by economic improvement and the need for efficiency [1] - 31% of large enterprises lose over one million annually due to excess costs and non-compliant orders, while over 60% of frequent business travelers waste 1.5 workdays each year due to cumbersome travel processes [1] Group 2: Company Initiatives - Didi Enterprise Edition's new one-stop travel product aims to provide a unified and comprehensive solution for various travel needs, including car rentals, flights, hotels, and train tickets [1] - The AI Xiaodi Travel Assistant and Management Assistant are based on 700,000 accumulated B-end service data and utilize an open-source Multi-Agent architecture [2] - The goal is to make business travel management precise and invisible, enhancing employee experience and management efficiency [2]

Core Insights - The fourth generation of the "MiniCPM" model, known as MiniCPM 4.0, has been released, featuring two parameter scales: 8B and 0.5B, achieving the best performance in its class [2][3] - MiniCPM 4.0-8B model utilizes a sparse attention mechanism, demonstrating performance comparable to Qwen-3-8B while requiring only 22% of the training cost [2][4] - The model achieves a remarkable inference speed of 600 Token/s, with a 220x acceleration in extreme scenarios, significantly enhancing long text processing capabilities [2][3] Performance and Architecture - MiniCPM 4.0 offers a 5x acceleration in long text inference speed compared to similar models like Qwen-3-8B and Llama-3-8B, with a maximum acceleration of 220x under memory-constrained conditions [3][4] - The model's architecture, InfLLMv2, reduces the sparsity from the industry standard of 40%-50% to just 5%, allowing for efficient long text calculations with only 1/10 of the computational load [4] - In terms of memory usage, MiniCPM 4.0-8B requires only 1/4 of the cache storage space compared to Qwen3-8B for 128K long text scenarios, indicating significant model compression and efficiency [4] Applications and Market Impact - Based on the 8B version, the company has fine-tuned two specific capability models for use as MCP Client and a research tool, MiniCPM4-Surve, which competes with Deep Research [5] - The MiniCPM series has achieved over 10 million downloads across all platforms, indicating strong market interest and adoption [5]

Core Insights - Meituan has officially launched its first AI Coding Agent product, NoCode, which allows users with no programming background to create websites and software tools through natural language interactions [1][2] - The NoCode tool integrates Meituan's self-developed LongCat model, which has a scale of hundreds of billions of parameters and performs comparably to leading models like GPT-4o and Claude-3.5-Sonnet in various benchmarks [1] - The product has been optimized for web front-end development, enabling efficient understanding and generation of HTML, CSS, and JavaScript code [1] Group 1 - NoCode was developed in just six months, originating from an internal "hackathon" event, and has already been widely used in various applications such as mini-games and ordering systems during its beta testing phase [2] - The official release of NoCode has further streamlined the development, testing, deployment, and bug-fixing processes, allowing users to complete complex tasks quickly without programming experience [2] - Approximately 50% of new code generated by Meituan employees is now produced by AI tools, with over 90% of engineering team members utilizing AI programming tools [2] Group 2 - The launch of NoCode aims to simplify programming tasks for small and medium-sized businesses, lowering the barriers to IT and digitalization [2] - Meituan plans to gradually open up more capabilities of its self-developed AI products in the future to promote technological inclusivity [2]

王兴表示，之所以不怕远去选择市场，是认为巴西是一个好市场，而且对于中国公司来讲，尤其是相对 靠谱的选择。因为很现实的情况，考虑的不光是那个市场有没有钱，还得考虑作为公司的身份，它是否 长期经营有保障的。从这一点来看，我们中国和巴西两个国家，保持着长期战略合作伙伴关系，这两国 领导人也频繁互访，进一步夯实的基础。中国企业在巴西发展应该充满了很好的环境。 新浪科技讯 6月10日上午消息，近日，美团股东大会在北京美团总部恒电大厦举行。在现场提问环节， 王兴回应了为何美团要大力开拓巴西市场。 责任编辑：屠欣怡 另一方面，王兴提到，巴西是一个大的经济体，是金砖最初的四国之一，按照地域算，巴西是全球第五 大，按人口算，巴西是全球第七大，按照经济体量GDP算的话，是在第九上下。但是它的外卖发展是不 错的，按目前外卖规模是第五名。所以巴西是很有潜力的市场。另一方面，巴西跟中国的经贸关系比较 稳定。所以我们看好巴西市场。但是另一方面我们不会特别着急，一是特别远，语言、文化都非常不一 样，另外巴西作为一个大国，它本身也有非常强的互联网公司。相对的好处，巴西总体开放，不管是欧 美的品牌，还是日本品牌，还有中国品牌，在那里都已经很多 ...

王兴表示：所以美团坚决反对这件事情。因为我们认为这个行业，虽然还有很大空间，但是毕竟发展了 十几年，中间经过了几轮竞争和调整，现在应该已经到了能够相对健康发展的阶段。如果再来一轮新的 以大额促销，以低于成本价促销为特征的冲击，对整个行业长期健康发展是不利的，而且我觉得和现在 强调各行各业反内卷的精神是相悖的，所以我们监管部门也会看到这种情况，也会出一些引导措施，美 团非常坚决反对内卷。（罗宁） 责任编辑：屠欣怡 新浪科技讯 6月10日上午消息，近日，美团股东大会在北京美团总部恒电大厦举行。谈及美团如何看待 当前的外卖平台竞争格局，王兴表示，美团是坚决反对内卷的。因为一个行业有更多玩家，有更多竞争 都是正常的，但是这个竞争到底是比较健康的，能够推动行业发展的竞争，还是不太健康的，很无效的 内卷，这个差别很大。如果大家真的关注这个市场，可以看到最近大家宣布单量上涨很快，但其中很大 比例是低价补贴导致的很低客单价的单，尤其是有很大供给弹性的奶茶和咖啡的单。就算有补贴，一个 人一天也不会吃五顿饭，也不会从两顿三顿变成五顿六顿。但原来不喝奶茶咖啡的，如果非常低价补 贴，甚至低到三、四块钱一杯，那可能他原来不喝的变成喝了 ...

Group 1 - The company expresses confidence in its long-term development despite increased competition in the food delivery platform market, attributing this to its early entry and established business foundation [1] - The company emphasizes the importance of both quantity and quality in its growth, particularly in food safety, and has taken measures to eliminate ghost restaurants and subsidize merchants for equipment and installation costs [2] - The company has launched initiatives to simplify marketing and address business pain points, welcoming more participants in the market while opposing internal competition, and remains optimistic about long-term success [3] Group 2 - The company has cleaned up over 16,000 ghost restaurants and has partnered with major telecom operators to promote transparency in food preparation through the "Internet + Bright Kitchen" initiative [2] - The company has opened 3,000 high-quality satellite stores in collaboration with 480 brands to enhance consumer trust and meet delivery demands in areas not fully covered by original stores [2]

新浪科技讯 6月10日上午消息，近日，美团股东大会在北京美团总部恒电大厦举行。在现场提问环节， 王兴回应了美团在AI领域方面的进展和策略。 王兴谈道，美团有两万多名研发人员，不光是程序员，包括其他的产品经理、设计师，做商业分析的人 员，他们用AI提高工作效率，这是对我们来讲近期非常重要的事情。在这里面，我们做了让原来不太 会编程的人能编东西，叫"No Code"。因为我们在整个过程中发现不光要帮原来已经在编程的人，提高 他的编程效率，提高AI自动生成代码的比例。更重要的事情是0到1，让原来不太会编程的人，不太擅 长编程的人，他现在自己通过AI也能够编，这样他有需求的时候，不需要找同事，需要排期自己去 弄，只要他有想法，AI能自动帮他生成代码，这样就能实现，原来不编程的人能编程，让他的能力得 到扩展。所以我们在内部研发了No Code，这个No Code已经在我们内部应用于所有的专业智能团队， 包括产品经理，包括UI设计师，包括我们做商业分析的同事，而且也包括HR和财务人员，取得很不错 的效果。所以前段时间我们也把No Code向公众免费开放，大家可以尝试一下。现在我们统计到有1， 680个通过No Code的 ...