130. 张月光创业两年首次访谈:妙鸭不是AI Native产品、流程到上下文设计、One Way Door和乙女游戏
130. 张月光创业两年首次访谈:妙鸭不是AI Native产品、流程到上下文设计、One Way Door和乙女游戏
Podcast3 hr 14 min
Listen to Episode
Note: AI-generated summary based on third-party content. Not financial advice. Read more.
Quick Insights

The next major investment opportunity is in AI Agents, which are poised to become the next computing platform by automating complex tasks. Meta (META) is strategically positioned to capitalize on this trend, as validated by its smart acquisition of an AI agent startup. Conversely, the rise of specialized AI agents presents a significant disruption risk to established software companies like Adobe (ADBE). When evaluating opportunities, prioritize companies building true "AI Native" products that create entirely new user experiences rather than just adding features. In the near term, watch the gaming sector for companies using AI to innovate on dynamic character development and storytelling.

Detailed Analysis

AI Sector & "AI Native" Theme

  • The podcast guest, Zhang Yueguang (creator of the viral AI app MiaoYa), draws a critical distinction between simply using AI technology and building true "AI Native" products.
    • He describes his own creation, MiaoYa, as a successful application of AI to an existing user need (creating a high-quality profile picture) but clarifies it is not an AI Native product because it follows a fixed, predefined workflow.
    • A true AI Native product is defined as having an open-ended core experience built around the generative and contextual capabilities of AI. The product design focus shifts from traditional UI/UX to what the guest calls "Context Engineering"—designing the hidden context and prompts that guide the AI to produce the desired outcome.
    • The guest is bullish that this fundamental shift in product design creates significant opportunities for new startups to compete with established tech giants ("大厂").

Takeaways

  • Investment Thesis: The next wave of disruptive tech companies may not be the incumbents who are just adding AI features, but rather new companies building true AI Native products from the ground up.
  • What to Look For: When evaluating an AI company, investors should ask if the product creates a fundamentally new, open-ended user experience. A focus on "Context Engineering" and creating a superior, almost magical, final result is a key indicator of a potentially disruptive AI Native approach.

AI Agents

  • The guest expresses a strong bullish sentiment on AI Agents, believing they will become the next major computing platform, fundamentally changing how we interact with technology.
    • He contrasts the current internet paradigm of "killing time" (e.g., scrolling social media) with the agent paradigm of "saving time" or enhancing a user's capabilities.
    • His new startup, Docky, is focused on building an agent, starting with a tool to automate the creation of presentations (PPTs).
    • The long-term vision is for agents to assist with major life decisions (e.g., buying a car or house), which would give them immense economic power by positioning them at the center of high-value transactions.

Takeaways

  • Bullish Sentiment: The discussion frames the AI Agent space as a primary frontier for future technological growth and disruption.
  • Investment Focus: Investors should pay close attention to companies developing agent-based products. These may start with narrow, specific applications but often have a broader vision to become autonomous personal assistants.
  • Key Differentiator: A successful agent is more than just a chatbot. It's a tool that understands complex context, can take actions on the user's behalf, and reliably accomplishes multi-step goals.

Meta (META)

  • The discussion highlights Meta's acquisition of the company Minus, an early player in the AI agent space.
    • The guest views this as a "smart" strategic move by Meta to secure a position in what he sees as a critical future market.
    • He describes the acquisition price as "not expensive" and a "good trade" for Meta, suggesting that the future value of a successful agent platform is far greater than what was paid.

Takeaways

  • Bullish Signal for META: The expert commentary validates Meta's AI strategy, particularly its early move into the agent ecosystem. This can be viewed as a positive indicator for META's long-term growth prospects in AI.

Alibaba (BABA)

  • The guest created the viral AI photo app MiaoYa while working at Alibaba.
    • The project's success demonstrates that large corporations like Alibaba have the resources (the project used about 2 million in compute cards) and talent to innovate and launch hit products quickly.
    • However, the guest's decision to leave Alibaba to found his own AI agent startup highlights a potential challenge for large companies: retaining top talent who want to pursue high-risk, paradigm-shifting ideas that may not fit into the corporate structure.

Takeaways

  • Mixed Signal: Alibaba clearly has the capacity for AI innovation. However, investors should be aware of the risk of talent drain, as visionary product leaders may depart to build the next generation of AI-native companies. Investing in BABA offers exposure to its vast resources, but it may not be where the most disruptive "pure-play" AI innovation occurs.

AI in Gaming

  • The guest discusses the massive potential and current limitations of AI in the video game industry.
    • Long-Term Vision: A true "AI Native game"—where the world, characters, and story are all dynamically generated in real-time (an "AI Zelda")—is seen as a holy grail but is not yet technologically feasible.
    • Near-Term Opportunity: A more immediate opportunity lies in games where character interaction and relationships are central. The guest specifically mentions "Otome games" (female-oriented dating simulators), where AI can create much more dynamic and evolving relationships with non-player characters (NPCs), enhancing the core experience.
    • Risk Factor: The guest points out a key weakness in current AI companion products like Character.ai: the characters are static and do not change or grow over time, which limits their ability to form long-term, meaningful "relationships" with users and could impact retention.

Takeaways

  • Investment Timeline: True AI-native open-world games are a distant but potentially enormous opportunity. In the near term, investors should look for gaming companies that are cleverly using AI to innovate on character development and dynamic storytelling, as this is where the technology can have the most immediate impact.

SaaS Disruption (Adobe, Canva)

  • The guest's new company, Docky, is building an AI agent to create presentations, a task currently dominated by established Software-as-a-Service (SaaS) companies.
    • Incumbents like Adobe (ADBE) and Canva are mentioned as the traditional players in the creative and productivity software market.
    • The strategy for disruption is to focus on a specific, high-value task and deliver a vastly superior result through an AI-native workflow.
    • The guest emphasizes the "90 is equal to 0" principle: an AI tool is only truly disruptive if it can deliver a result that is 99-100% complete. If a user has to spend significant time fixing a 90% good result, the tool has failed to provide real value.

Takeaways

  • Disruption Theme: The rise of specialized AI agents poses a significant competitive threat to the business models of established SaaS giants like Adobe and Canva.
  • Evaluating AI Products: When assessing an AI tool, the most important factor is the quality and completeness of its final output. A product that can reliably deliver a near-perfect result is exponentially more valuable and disruptive than one that just offers assistance.
Ask about this postAnswers are grounded in this post's content.
Episode Description
过去两年,张月光仿佛消失了。 这位产品经理在2023年创造过妙鸭这样现象级AI产品,但在妙鸭走红后,仅3个月他便从阿里辞职。他说,这源于他独自在日本京都的鸭川边,坐了一夜,想清楚了许多重要的人生决定。 2024年初,张月光筹措资金,创业沐言智语。这家公司成为那年AI应用型公司里融资额最高的初创公司之一。 有人说,投资人在没有产品的情况下,对他“盲注”4轮,累计金额近三亿元人民币。也有人开玩笑,张月光是在拿着融资额的利息养公司。 过去两年,我和张月光见了很多次,他在尝试的产品形态每一次都有一些变化,他也并不掩饰自己的困惑,坦言的确有一段时间“看不懂”。转折来自半年前。 一位师承于互联网时代的产品经理,当他转向AI,会经历怎样的试错与求索?又要跨越怎样的认知鸿沟?——让我无比意外的是,张月光说,他后来才意识到,妙鸭根本不是AI Native的产品。 沐言智语现阶段正在做的有两款主力产品:AI乙女游戏《星眠》,和即将要推出的Agent产品Dokie。他说,他的探索期结束了,他终于找到那个他愿意集公司之重力下注的东西。 接下来是我对张月光的访谈。 期待2026年,我们和AI共同进步! (张月光坐在鸭川边思考人生的长夜) 02:00 嘉宾小传 2012年清华计算机毕业,大三发现自己卷不了Coding 2018年第一次失败的创业与产品(元音) 上次创业到最后非常困惑:不知道自己为什么要开始创业 我在字节的两段经历(时光相册、今日头条专项) 这次创业前的人生链路:SaaS公司>支付宝>字节>创业>阿里优酷 2023年国庆节的凌晨,在京都鸭川边坐着,决定了第二次创业 我的原则:“我想35岁-45岁,和同一群人做同一件事情” 16:56 在大厂打怪升级的游戏 27岁是字节3-2,30岁是阿里P9 大厂普遍有两种战役:外部战役和内部战役 坦白说,大厂游戏会让绝大部分人变成内部作战的人 AI提供了外部游戏的增量,否则只能打内部游戏 创业最好的理由是两个:1、你就想当老板;2、你特别想让一件事发生;但凡想发财、想成名,创业的期望收益都很差 长视频平台不是平台,就像电影院不是平台一样,说到底它是内容产品 2023年我开会说:“从今天开始关于优酷的各种会我都不参与了” 28:28 妙鸭不是一个AI Native产品,那什么是? 我喜欢One Way Door产品(单向门产品) 互联网时代的产品设计是面向流程的设计,好的产品比如微信是游乐园,但也是有限自由度 但AI Native产品非常开放,不仅是输入开放,输出也是开放的 从互联网产品设计到AI Native产品设计,有一个巨大的范式变化:从流程设计到上下文设计 传统互联网产品团队的协作方式在今天很有可能出现问题 42:04 设计AI产品究竟需要什么样的AI组织?决策边界在变模糊,Taste变得非常重要,从线形工作变成两段式 妙鸭和美图秀秀不在一桌,它本质是写真生意 我们发现很难让一个模型做到真、像、美,于是我们让三个模型同时做 最多时用了接近2万张推理卡 妙鸭添加了我人生的幸福时刻:和同事在亮马桥喝酒那天,日收入破了100万 2023年7月到10月忙得屁滚尿流,10月休假决定辞职,一切戛然而止 01:01:16 出发第二次创业啦 这代AI技术应该是“创造AI人口”,而不是“创造AI服务” AI会被连接到整个人类网络里来 今天的专业技能正在贬值,多元化的品味、博学程度决定团队上限 “人应该负责Will,人应该负责Skill” 2023年作为融资额最高的AI应用公司之一,投资人为什么在没有产品的情况下,“盲注”? 行业里的人开玩笑说,“我们天天在花利息” 01:14:29 上路后,产品的多次选择和变化 一条确定的线路:以交互为核心的AI乙女游戏(星眠) 一条不确定的线路:花了1年摸索,做了三四个小产品,比如播客、信息工具、图像工具 说说我们为什么之前做了一个播客产品,然后放弃了? 新的内容平台的唯一机会:新媒介、新交互,产生新平台 Sora绝不可能成为新平台,但是ChatGPT可以 我觉得Chatbot的商业空间是难以想象的大,可能是人类有史以来从没见过的商业模式——它背后藏着长程的决策控制的能力 抖音式的短视频交互剥夺了你的信息自主权,Chatbot有机会剥夺你的自主决策权 它可以控制你,那么它可以拿很高的Take Rate(抽佣率) 02:01:21 从模型范式过渡到Agent范式 让Agent做到更多你完全做不到的事情 Manus是一个“比硅谷更硅谷的童话故事”,Meta买到了一个范式的代表 我不想做Manus这样的Agent,即将发布新产品:Dokie 我很在意Latency(延迟),希望持续协作 为什么Dokie选择PPT这么窄的赛道?——“别说the first了,我们都开始the last了” 我的产品哲学:不是the first,90分等于0分,没有人觉得这是AI做的,做出One Way Door/单向门体验 回答天问:怎么看待和模型公司的竞争?(应用公司都在用Claude的tool call) 星眠做的不是我的AI朋友,Dokie做的是一个突破我能力边界的AI朋友 【更多信息】 免责声明:本内容不作为投资建议。
About 张小珺Jùn|商业访谈录
张小珺Jùn|商业访谈录

张小珺Jùn|商业访谈录

By 张小珺

努力做中国最优质的科技、商业访谈。 张小珺:财经作者,写作中国商业深度报道,范围包括AI、科技巨头、风险投资和知名人物,也是播客《张小珺Jùn | 商业访谈录》制作人。 如果我的访谈能陪你走一段孤独的未知的路,也许有一天可以离目的地更近一点,我就很温暖:)