AI替身 · 不用学习 AI 的个人基座 · 邀请制私测 AI替身 · PERSONAL AI WITH NO LEARNING CURVE · PRIVATE ALPHA
AI替身在你明确授权的范围内，通过手机、电脑以及后续接入的穿戴设备、智能眼镜、录音设备和具身智能持续理解你的目标、处境、偏好与所见所闻，主动发现问题与机会。你不必学习提示词、工具或 Agent，也能直接用上 AI 的能力。 With your explicit permission, AI替身 learns your goals, context, preferences and lived signals through phones, computers and, over time, wearables, smart glasses, recording devices and embodied AI. It proactively surfaces problems and opportunities without requiring you to learn prompts, tool stacks or Agents.
当前能力：主动建议 · Windows / Android 已运行 · Agent 执行尚未开放 Available now: proactive advice · Windows / Android running · Agent execution not yet available
你的目标YOUR GOAL
为明天的投资人会议完成市场证据页Finish the market evidence page for tomorrow’s investor meeting
所见SIGNAL
会议还有 18 小时，第 4 页只有未经核验的市场数字。The meeting is in 18 hours. Slide 4 contains unverified market figures.
判断JUDGMENT
先对齐来源与统计口径，再写入材料，避免把重叠市场相加。Align sources and definitions before writing, so overlapping markets are not added together.
当前功能示意：发现问题并主动建议 · 执行能力属于后续路线 Current capability: proactive discovery and advice · Execution remains on the roadmap
01
提示词要学，工具要学，Agent 也要学；而最需要 AI 的老板、高管和专业工作者，往往最没有时间学习。AI替身把关系反过来：不是让每个人成为 AI 专家，而是让 AI 在授权范围内先学会你。 People are expected to learn prompts, tools and Agents, while leaders and professionals responsible for the highest-value decisions often have the least time. AI替身 reverses that relationship: instead of making everyone an AI expert, let AI learn the person within clear permissions.
02
多端不是只有手机和电脑。手机、电脑、穿戴设备、智能眼镜、录音及其他传感设备，以及未来的具身智能，都会在你逐项授权后成为信息感知端，收集操作、工作内容、身体状态、情绪相关信号和所见所闻，共同形成一份由你控制、可查看、可纠正、可撤回的个人上下文。 Multi-device means more than phones and computers. With your permission for each source, phones, computers, wearables, smart glasses, recording and other sensing devices, and future embodied AI can all become sensing inputs—capturing actions, work context, physical state, emotion-related signals, and what you see and hear—into one context you can review, correct and withdraw.
03
当前功能只做两件事：从你授权的信息中主动发现问题，并生成贴合你的建议。下一阶段才是在确认下调度 Agent、操作软件、沟通、购物与支付。 The current product does two things: proactively identifies issues from authorized context and generates advice that fits you. Agent coordination, software operation, communication, shopping and payments come later—with confirmation.
结合你的目标与当下处境，在你还没想到问 AI 时，先识别值得处理的风险、偏差和机会。Use your goals and current context to surface risks, gaps and opportunities before you think to ask an AI.
理解完整上下文后起草消息、协调日程和跟进任务；任何对外发送都先展示给你确认。Draft messages, coordinate schedules and follow up with full context; show you anything before it is sent.
根据既有偏好、预算与真实用途完成比较和准备；到下单、付款或订阅时把决定交还给你。Compare and prepare choices using your preferences, budget and real use case; return control for ordering, payment or subscription.
在清晰权限下协调互联网服务；具身智能成熟后，它既通过视觉、听觉和环境传感补充现实信息，也在同一套身份、权限与确认机制下代表你完成现实任务。Coordinate internet services within clear permissions. As embodied AI matures, it can both add real-world context through vision, audio and environmental sensing, and complete physical tasks under the same identity, permission and confirmation layer.
04
搜索帮助人找信息，商业平台帮助人购物，聊天推动沟通，支付完成交易。AI替身的长期方向，是先理解个人意图，再协调搜索、购物、娱乐、本地服务与专业服务完成结果，从而重构入口、流量、佣金与交易价值的分配。 Search finds information, commerce enables shopping, messaging drives communication, and payments close transactions. The long-term direction of AI替身 is to understand personal intent first, then coordinate search, shopping, entertainment, local and professional services toward an outcome—reshaping how discovery, traffic, commissions and transaction value are distributed.
搜索与发现SEARCH & DISCOVERY
$224.5B
Alphabet 2025 年 Google Search & other 收入。它是单家公司收入，不等于全球搜索市场。Alphabet’s 2025 Google Search & other revenue. This is one company’s revenue, not the global search market.
来源：Alphabet 2025 10-KSource: Alphabet 2025 10-K数字商业与个性化购物DIGITAL COMMERCE & PERSONAL SHOPPING
$28T
45 个经济体 2024 年企业电商销售额，覆盖约全球四分之三 GDP；包含 B2B 与 B2C。2024 business e-commerce sales across 45 economies representing roughly three quarters of global GDP; includes B2B and B2C.
来源：UNCTADSource: UNCTAD聊天、商业与 AgentMESSAGING, COMMERCE & AGENTS
1B+ / DAY
Meta 2026 年披露，WhatsApp、Messenger 与 Instagram 每天产生超过 10 亿次企业活跃会话。Meta reported in 2026 that WhatsApp, Messenger and Instagram generate more than one billion active business conversations each day.
来源：MetaSource: Meta数字支付DIGITAL PAYMENTS
$18.7T
2024 年数字支付方式承载的线上与线下消费额。它是支付流水，与电商统计高度重叠。Online and in-store consumer spend carried by digital payment methods in 2024. This is payment flow and overlaps heavily with e-commerce.
来源：Worldpay GPR 2025Source: Worldpay GPR 2025口径说明：以上是 AI替身可能重构的相邻价值池，统计年份和定义不同、彼此高度重叠，不能相加，也不是 AI替身可获得的收入。 Method note: these are adjacent value pools AI替身 may reshape. They use different years and definitions, overlap heavily, cannot be added together, and are not revenue available to AI替身.
创始人情景判断：若零学习门槛、持续真实价值与用户信任都得到验证，这类个人 AI 长期触达全球 50% 以上人口，并产生千亿美元至数十万亿美元级经济影响，具有想象空间。“社会产出提高 10—100 倍”是极端成功情景下的思想实验，不是公司估值、收入、采用率、GDP 或投资回报预测；重构巨头的部分功能与入口，也不等于必然替代任何具体公司。 Founder scenario: if near-zero learning effort, sustained real value and user trust are all demonstrated, personal AI systems could conceivably reach more than 50% of the global population and influence hundreds of billions to tens of trillions of dollars in economic activity. “10–100× social output” is an extreme-success thought experiment—not a company valuation, revenue, adoption, GDP or investment-return forecast. Reshaping functions and entry points does not imply the inevitable replacement of any company.
05
蒸汽动力通过机械化生产释放价值，电力通过流水线与大规模生产进入日常生活。我们的产品判断是：基础模型、编程系统和 Agent 正在快速成熟，缺的是一个长期围绕单个人工作、把分散能力整合起来的系统。AI替身要成为这个个人应用层——今天连接互联网，未来从物理设备和机器人接收经授权的现实感知，并经确认通过它们执行现实任务。2024 年全球已有 466.4 万台工业机器人在运行。 Steam power created value through mechanized production; electricity reached daily life through assembly lines and mass production. Our product thesis is that foundation models, coding systems and Agents are advancing quickly, while a system continuously organized around one person is still missing. AI替身 aims to become that personal application layer—connected to internet services today and, tomorrow, receiving permissioned real-world sensing from physical devices and robots while using them to perform tasks only after confirmation. There were 4.664 million industrial robots operating worldwide in 2024.
06
AI替身的方向不是让人服从 AI，而是扩展每个人的理解、判断与行动能力。个人上下文应当可以查看、纠正和撤回；涉及身份、沟通、支付或现实设备的动作必须由用户确认。把重复执行交给 AI，把重要选择、关系与创造留给人。 AI替身 is not intended to make people subordinate to AI, but to extend each person’s understanding, judgment and action. Personal context should be reviewable, correctable and revocable; actions involving identity, communication, payments or physical devices must require user confirmation. Let AI handle repetition while important choices, relationships and creation remain human.
可见界面、通知、文件、语音与云分析分别由用户开启，并可随时暂停。Visible UI, notifications, files, voice and cloud analysis are independently enabled and can be paused.
密码框、安全窗口和其他应用的私有数据库不属于 AI替身的读取范围。Password fields, secure windows and other apps’ private databases remain outside the reach of AI替身.
付款、发送消息、修改账户、公开发布和删除等外部或不可逆操作，不因“Agent”而跳过确认。Payments, messages, account changes, public publishing, deletion and other external or irreversible actions never bypass confirmation merely because an agent is involved.
用户能看到为何行动、使用了什么工具、结果是否回读成功，并用真实结果纠正下一次。Users can see why an action was proposed, which tool was used, whether the result was verified, and correct the next attempt with real outcomes.
07
AI替身仍是邀请制私测。当前只聚焦授权数据、主动发现问题与生成个人建议；聊天、购物、支付、软件操作和机器人控制都属于后续路线，不会写成已经完成的能力。 AI替身 remains an invitation-only private alpha. The current product focuses only on authorized context, proactive discovery and personal advice. Messaging, shopping, payments, software operation and robot control remain on the roadmap and are not presented as shipped capabilities.
产品进度PRODUCT PROGRESS
发布原则：每项能力都以真实设备、真实权限和执行后回读为准。Release principle: every capability must be proven on real devices, with real permissions and post-action read-back.
08
宇瀚（深圳）人工智能有限公司最初为高频广告运营构建自动决策、执行与风险控制系统。我们由此看到一个更大的机会：把只服务单一业务的自动化，变成长期围绕每个人工作、无需学习就能使用全部 AI 能力的个人基座。AI替身由此诞生。 Yuhan (Shenzhen) Artificial Intelligence Co., Ltd. began by building automated decision, execution and risk-control systems for high-frequency advertising operations. That work revealed a larger opportunity: turn automation built for one workflow into a personal foundation that works continuously around each person and makes AI useful without a learning burden. AI替身 grew from that conviction.
我们的目标，是让 AI 先学会人，再把全部能力带到人面前：今天从理解与建议开始，未来走向代表行动与结果负责。 Our goal is for AI to learn the person first, then bring its capabilities to them—starting with understanding and advice today, and moving toward accountable action tomorrow.
体验 · 共建 · 推广 · 投资TRY · BUILD · DISTRIBUTE · INVEST
你可以申请体验 AI替身当前的建议能力，也可以与我们共同建设产品、数据安全、行业场景和推广渠道，或洽谈投资合作。具体分成、权益与股权以双方书面协议及适用法律为准；本页不构成募资邀约或投资回报承诺。 Apply to experience the current advice capabilities of AI替身, or work with us on product development, data safety, industry use cases, distribution and investment. Revenue sharing, rights and equity are subject to written agreements and applicable law; this page is not a public securities offering or a promise of investment returns.
09
选择你的需求，留下必要信息。我们会用它评估演示、投资交流或私测申请。Choose what you need and share only the essentials. We will use them to assess a demo, investor conversation or alpha application.
请勿提交密码、付款信息或其他敏感数据。Do not submit passwords, payment details or other sensitive data.