In the age of AI, there’s a new currency for success, and it’s not just about what you know. It’s about how fast you can turn that knowledge into action. This is the “Knowledge-to-Action Loop,” and AI is the bridge that makes it happen instantly. This principle is not new — it echoes the ancient Chinese wisdom of 知行合一 (zhī xíng hé yī), the unity of knowledge and action.

1. Vibe Coding: From Idea to Prototype in Minutes

Every experienced professional knows the pain: you want a small tool or workflow fix, but the request disappears into the IT backlog. By the time it comes out, it’s either irrelevant or unrecognizable.

That’s the old world: knowledge (the idea) separated from action (the result).

The concept of Vibe Coding is the ultimate micro-example of the Knowledge-to-Action loop in practice.

It’s not about writing code; it’s about sketching with it. You toss out an idea, and an AI tool generates a first-draft prototype. Want changes? It adapts instantly.

The process is a continuous, rapid-fire cycle of Idea → Feedback → Iteration → Usable result.

• Traditional coding: write the “sheet music” (logic) for days, play it for weeks, restart if a note is wrong.
• Vibe coding: pick up the “guitar” (AI tools) and jam — mistakes fixed on the fly, usable output in minutes.

This is knowing and doing converging in real time.

2. MBZUAI: The Institutional Blueprint for “Knowing-Doing”

While Vibe Coding is personal, some institutions are building this philosophy into their DNA. The Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) is a prime example. Founded in Abu Dhabi in 2019, it is the world’s first university dedicated entirely to AI — not to produce theorists, but leaders solving real-world problems.

Their president, Eric Xing, is the living embodiment of this principle. His career isn’t siloed; it’s a seamless loop:

• Academic “Knowing”: A dual Ph.D. in Molecular Biology and Computer Science and CMU professor, mastering the theoretical underpinnings of AI.
• Industry “Doing”: He co-founded Petuum, a company that scaled distributed machine learning from the lab to the enterprise, earning a $93M Series B from SoftBank. Then, he launched GenBio AI to use AI to build “digital organisms” that can simulate DNA and proteins, turning his academic knowledge into a tool for biotech and pharma.

Research is the “knowing,” and entrepreneurship is the “doing.” He treats AI not as abstract equations, but as a converter that turns theory into solutions.

A Local Problem, A Real Fix

A perfect example is MBZUAI’s work on deepfake detection for the Middle East. They saw a unique, local problem — the widespread use of “Arabish” (a mix of Arabic and English) in daily conversation.

MBZUAI spotted this blind spot for deepfake detection systems:

• Knowing: Human detection accuracy was just 60%; existing AI accuracy dropped by 35% in mixed-language cases.
• Doing: Built ArEnAV, a 765-hour bilingual audio-visual dataset. This became the global benchmark for bilingual deepfake detection.
• Value: Media outlets and fact-checkers can now reliably flag fakes in Arabic-English content.

Their paper title says it all: “Tell Me Habibi, Is It Real or Fake?” It’s not about the tech; it’s about solving a local, human problem.

3. Young Founders: Age No Longer a Barrier

The traditional model of entrepreneurship required years of experience, a robust network, and substantial funding. AI has levelled the playing field, introducing a new form of leverage beyond human resources and capital. Today, the core competitive advantage is no longer what you have, but how fast you can execute.

Look at the young founders breaking through:

• Brenden Foody: Launched Mercor, an AI-powered recruitment platform, at just 19. AI handled the candidate matching and resume analysis, allowing him to build a prototype and secure major funding by age 22.
• Adam Guild: Started young, spotted restaurant owners’ pain — no digital capability. With AI, he built tools to automate marketing and operations, scaling Owner.com to unicorn status by 25.

The common thread? Not just youth, but the ability to turn ideas into working products fast with AI.

4. The Future Belongs to Creators of “Knowledge-Action Unity”

As Auguste Rodin famously said, “The world is not lacking in beauty, but in discovering eyes.” In the AI era, the same holds for technology: the world isn’t lacking in tools, but in people who can wield them to solve problems.

AI itself is merely a tool. Its true value isn’t inherent in the technology, but in the skill of the user to leverage it. Consider the vast potential of AI tools like ChatGPT: while some may use it for casual purposes like fortune-telling, true innovators will harness it for coding, building systems, and creating products.

The fundamental survival logic in the AI age is this: those who can rapidly translate “knowing” into “doing” with AI will remain competitive. Your degree of “knowledge-action unity” will ultimately dictate your standing and impact in this new landscape.

先抛结论：AI 时代的竞争，本质是 “知行合一” 的竞争 —— 而 AI，正是打通 “知道” 与 “做到” 的关键桥梁。过去我们被 “想法到落地” 的鸿沟卡住，现在 AI 把这条沟填成了平路。

一、Vibe Coding：从 “空想” 到 “落地” 的即时闭环

老职场人都懂一个痛点：想做个小工具、改个功能，把需求发给 IT 部门，大概率就掉进了 “流程黑洞”—— 等排期、等资源、等反馈，最后要么不了了之，要么出来的东西早已偏离初衷。这就是典型的 “知”（想法）与 “行”（落地）脱节。

AI 彻底改写了这个逻辑，Vibe Coding（氛围式编程） 就是最好的例子。它的核心不是 “写代码”，而是 “像画草图一样玩代码”：你抛出想法，AI 先出第一版原型；你需要调整，AI 秒改；你再提要求，AI 继续调整 —— 整个过程没有等待，只有 “想法→反馈→迭代” 的即时循环，直到拿到你想要的结果。

这正是 “知行合一” 的微观落地：

• 传统编码：先花 3 天写 “乐谱”（代码逻辑），再花 1 周 “演奏”（调试运行），错一个音符就得重来；
• Vibe Coding：直接拿起 “吉他”（AI 工具）即兴弹，弹错了立刻调整，5 分钟就能出一段 “能听的旋律”（可用的原型）。

知与行，在一次次对话中完成闭环。AI 在这里的角色，是 “即时执行者”：它把你脑子里的 “知”，瞬间转化为可触摸的 “行”，再通过你的反馈快速优化，让 “想” 和 “做” 变成同一件事。

这种“知行合一”的实践，正在全球最前沿的AI组织和教育机构中成为现实。

二、MBZUAI：机构级 “知行合一” 的范本

如果说 Vibe Coding 是个人层面的 “知行闭环”，那穆罕默德・本・扎耶德人工智能大学（MBZUAI） 就是机构级的标杆 ——2019年成立于阿布扎比，是全球首所专注于AI的大学。它的使命不是培养“理论家”，而是打造能解决真实世界问题的AI领袖。

而它的掌舵人，正是“知行合一”的典范。

（一）、校长邢波：从 “双博士” 到 “双创业者” 的知行典范

MBZUAI 校长邢波（Eric Xing），本身就是 “知行合一” 的活教材。他的履历里没有 “割裂感”：

• 学术端（知）：清华物理系本科，分子生物学 + 计算机双博士，卡内基梅隆大学CMU 正教授 —— 吃透了 AI 的底层逻辑，还懂生物、医疗的真实需求；
• 产业端（行）：2016 年创 Petuum，把 “大规模分布式机器学习框架” 从实验室推到产业，软银领投 9300 万美元 B 轮，成了 世界经济论坛 (WEF) 认证的 “技术先锋”；2024 年再创 GenBio AI，用 AI 构建 “数字生命体（AIDO）”，直接模拟 DNA、蛋白质功能 —— 把学术知识，变成能解决医疗、制药痛点的工具。

刑波校长的逻辑很清晰：AI 不是论文里的公式，而是要落地的解决方案。学术研究是 “知”，创业是 “行”，AI 就是中间的转换器。在知与行之间不断往返，每一次循环都不断放大价值。

（二）、 阿英深度伪造检测：解决 “别人看不见的痛点”

MBZUAI 不仅在学术与产业结合上树立标杆，也直面真实世界的复杂问题。例如，中东地区广泛存在阿拉伯语与英语混用的现象（如“Habibi, come to Dubai”），一句阿拉伯语+英语的混合表达，自然流畅。但对大多数深伪（deepfake）检测系统来说，这是“无法识别的噪音”。

MBZUAI 演示了从 “知”（发现痛点）到 “行”（用 AI 解决）的完整落地：

• 第一步（知）：团队发现 “多语言伪造检测是空白”，且人类识别准确率仅 60%，现有系统准确率只有 35%；
• 第二步（行）：联合蒙纳士大学建了ArEnAV 数据集— 765 小时真实 “阿英混说” 语音视频，覆盖方言切换、语言跳转，成了全球最大的双语伪造检测基准；
• 第三步（价值）：这份数据集直接给媒体、事实核查机构用，补上了中东 “反虚假信息” 的短板。

他们的论文标题很有意思：《Tell Me Habibi, Is It Real or Fake?》（亲爱的，这是真的还是假的？）—— 没有空谈技术，而是直击当地人的真实焦虑。这就是 AI 时代的 “知行合一”。

三、AI 平权：年龄不是问题，“知行闭环速度” 才是

过去想创业，得攒经验、攒资源、攒团队 —— 年轻人的想法再棒，也会被 “没资源落地” 卡住。但 AI 彻底撕了 “年龄门槛” 这张纸：现在的核心竞争力，不是 “你有多少经验”，而是“能否有效使用AI”。

我们看到越来越多年轻创始人脱颖而出：

• Brenden Foody（19 岁创业）：想做 “AI 驱动的招聘筛选工具”，不用自己搭复杂算法 ——AI 帮他搞定候选人匹配、简历分析，快速做出 Mercor 的原型。22岁领导公司完成重大融资。
• Adam Guild（最早13岁开始创业）：20岁左右瞄准独立餐厅的痛点 —— 没能力做数字化运营。他用 AI 搭了Owner.com，并在25岁时将其发展为估值超过10亿美元的独角兽企业。

他们的共同点是什么？不仅是 “年轻”，还都能 “用 AI 把想法快速变成产品”。

四、未来属于“知行合一”的创造者

雕塑家罗丹说，“世界上并不缺少美，而是缺少发现美的眼睛。” 在AI时代，世界上并不缺少技术，而是缺少利用技术解决问题的人。

AI是工具，工具本身不产生价值，但善于使用工具的人可以。

就像 DeepSeek，很多人用它算命看星座，但真正的创造者，会用它写代码、建系统、做产品。

所以 AI 时代的生存逻辑就是：谁能把 “知道” 通过 AI 快速变成 “做到”，谁就能留在牌桌上。知行合一的程度，就是你在 AI 时代的生存高度。

视频版

一、AIvilization 的核心： “涌现社会”

最近，我深度体验了香港科技大学（HKUST）在 2025 年 8 月 19 日刚发布的一个新项目，叫 “AIvilization”。

光看名字就能猜个大概 —— 它是 “AI（人工智能）” 和 “Civilization（文明）” 拼起来的，中文可以叫 “AI 文明”。据说是目前全球最大的 “AI 多智能体社会模拟沙盒”，听着有点绕？其实你可以把它理解成一个 “10 万 AI 的大型在线生存游戏”—— 但它真正厉害的，并不仅仅是AI 数量多，而是藏在背后的 “涌现社会” 逻辑。

之前 Meta、谷歌也做过 AI 模拟，但大多是让 AI 完成指定任务，比如 “合作搬东西”或“回答问题”等等；而 AIvilization 的 10 万 AI 不一样，它们装了香港科大自己研发的 “动态因果交互算法”。开发者没有提前写好 “谁该管事儿”“怎么交易” 这些规则，AI 只知道最基础的两件事：“得找资源活下去”和“跟别的 AI 打交道时，选合作还是竞争”。但就靠这点简单逻辑，它们在互动中居然自己 “搭” 出了复杂的社会结构 —— 有的专门负责干活（分工），有的拿东西换东西（交易），甚至还形成了小圈子（社群）。

这种现象叫 “涌现”，其实在自然界很常见：比如一只蚂蚁只会闻着气味搬食物，可成群蚂蚁凑在一起，不用谁指挥，就能造出分岔特别精密的蚁穴。AIvilization 里的 AI 也是这样，没有程序员鱼线编好的 “剧本”，完全是 AI 靠简单互动，自己 “长” 出来的秩序。

二、镜像世界预演

正如凯文・凯利（Kevin Kelly,《连线》创始主编）对未来 25 年的预判，这场实验更像一次 “镜像世界预演”。

他指出，AI 是 “镜像世界的无形基础设施”：到 2049 年，智能眼镜将取代手机，数十亿人会时刻穿梭于 “现实物理层 + 虚拟数字层” 的叠加态。

而 AIvilization 的价值，正在于提前叩问镜像世界的核心命题：当身份可虚拟、场景可模拟、眼见不再为实，作为社会协作基石的 “信任”，该如何建立？

三、我在沙盒中的数字分身

我在AIvilization沙盒里复刻了一个完整的 “数字分身”：不仅匹配成长背景、价值观，还导入了我的 MBTI 测试数据，甚至包括我对 “失败的容忍阈值”。

她首先扎进果园。

我问她： “摘苹果除了果腹还能做什么”，

她答： “分给附近没找到食物的智能体，交朋友”；

我进一步提议： “做点更挣钱的事”，

她却坚持： “先把果园照料好 —— 悉心付出的过程，本身就是让其他智能体信任我的方式”。

四、数字时代的信任构建

在算力驱动、效率至上的沙盒里，“她” 的选择像一记警钟。现实中，很多人连 3 分钟的视频都很难耐心地看完，急功近利、趋易避难成了商业决策与个人选择的常态。

但数字世界的底层逻辑，恰恰是信任 —— 国与国之间的协作、民众与政府的共识、消费者与企业的联结，甚至 “事物真伪的验证方式”，都需要重新构建。

AIvilization 的 “果园逻辑” 恰恰揭示：数字世界的底层规则与物理世界不同 —— 信任不是靠顶层协议强制约束，也不是靠短期利益快速换取，而是像照料果树一样，在持续输出价值（如稳定提供苹果）、传递正向互动（如主动分享）中自然生长。这与 “涌现社会” 的逻辑完全契合：复杂的信任体系从不是设计者规划的结果，而是无数个体在互动中积累正向反馈的 “涌现产物”。

五、数字人生报告

实验结束后，每位参与者会收到一份 “数字人生报告”：不仅涵盖分身的财富、工作成就、生活满意度、技能增长、社会关系五大维度数据，还会拆解其与其他 AI 智能体的交互模式 —— 比如 “我的分身” 在社群中是 “资源提供者” 还是 “协作发起者”，其信任度在不同社群中的评分差异等。

我格外期待这份报告，想看看这个 “复刻版的我”，在无现实规则束缚的 “涌现社会” 里，能参与构建出怎样的微小信任网络。

六、邀你共探社会范式转移

本质上，AIvilization 的实验不是 “AI 玩游戏”，而是一次社会范式转移的预演：我们即将迎来的不只是 “虚实叠加” 的下一代互联网，更是 “人机共生、规则自演化” 的新型社会结构。

技术能快速迭代算力、优化算法，但文明走向未来的关键，或许藏在 “果园逻辑” 里 —— 如何守护人性中的耐心、协作与信任，让这些 “非技术属性” 成为数字社会的底层支撑。

亲身参与沙盒的朋友，欢迎 “萃有集” 微信公众号，我们一起共同挖掘这场实验背后的未来信号。

视频版

When AI Societies Write Their Own Rules

On August 19, 2025, the Hong Kong University of Science and Technology (HKUST) launched AIvilization — the world’s largest AI multi-agent social simulation. Think of it as 100,000 AI agents dropped into a digital world with no laws, no governments, and no economy. Just basic instincts: acquire resources, decide whether to cooperate or compete, and adapt to changing conditions.

Unlike earlier AI simulations by Meta or Google, AIvilization uses HKUST’s dynamic causal interaction algorithm. This means rules aren’t designed top-down — they emerge from the bottom up.

A Rehearsal for the Mirror World

This experiment is a dress rehearsal for the future that Wired’s Kevin Kelly has long envisioned: the “Mirror World.” He describes AI as the “invisible infrastructure” for a new reality where, by 2049, smart glasses will have replaced our phones, plunging billions of us into a constant blend of the physical and digital worlds.

This is precisely where AIvilization proves its worth. It forces us to grapple with the fundamental question of the coming age: In a world where identity can be forged, reality can be simulated, and seeing is no longer believing, how do we establish trust — the very cornerstone of social collaboration?

My Digital Twin in the Sandbox

I created a complete “digital twin” in the sandbox, programming it with my background, values, MBTI profile, and even my specific threshold for failure. Her first action was to begin working in an orchard.

When I asked about the purpose of picking apples beyond simple sustenance, she responded, “To distribute them to nearby agents who lack food and to build alliances.”

I then proposed a more profitable venture, but she held her ground. “My priority is to care for this orchard,” she stated. “The very process of contributing with consistent care is how I earn the trust of others.”

Cultivating Trust in a Digital World

In a world driven by computing power and efficiency, her choices stand as a stark warning. Many people struggle to watch a three-minute video, and a focus on quick wins and easy solutions has become the norm for both business and personal decisions. Yet, the foundational logic of the digital world is trust. Collaboration between nations, consensus between the public and government, connections between consumers and businesses, and even how we verify truth, all require a new framework.

AIvilization’s “orchard logic” reveals a fundamental difference between the digital and physical worlds: trust isn’t a top-down mandate or something you can quickly earn with short-term gains. Instead, it grows organically, much like tending to fruit trees. It’s cultivated by consistently providing value (like offering a steady supply of apples) and fostering positive interactions (like proactively sharing). This aligns perfectly with the concept of an emergent society, where a complex system of trust isn’t the result of a single designer’s plan. It’s an emergent product of countless individuals building positive feedback through their interactions.

My Digital Twin’s Emergent Trust Network

After the experiment, participants will receive a “Digital Life Report.” This report will cover not only five key metrics for their digital twin — wealth, career achievements, life satisfaction, skill growth, and social relationships — but also provide insights into their overall well-being. It will also break down their twins’ interaction patterns with other AI agents. For example, it will show whether “my twin” was a “resource provider” or a “collaboration initiator” within a community and how their trust rating varied across different groups.

I am particularly excited about this report. I want to explore the potential of a small-scale trust network that this “replicated me” can help build in an emergent society, free from the constraints of real-world rules.

Beyond AI Gaming: A Rehearsal for a New Society

At its core, the AIvilization experiment isn’t just “AI playing games” — it’s a rehearsal for a paradigm shift. The future we are moving toward isn’t simply the next generation of the internet with virtual and real worlds layered on top of each other. It’s a new societal structure where humans and machines coexist and rules evolve on their own.

While technology can rapidly iterate on computing power and optimize algorithms, the key to civilization’s future might be hidden in the “orchard logic.” It’s about preserving human qualities like patience, collaboration, and trust, and letting these “non-technical attributes” serve as the foundational support for a digital society.

If you’d like to participate in this experiment, please send a private message to receive an invitation code. Let’s explore the future signals behind this experiment together.

我们身处一个创新不息的世界 ——AI 突破、量子计算，日新月异。我们的父母、祖父母、叔伯姨婶 —— 他们中的许多人亲历了人类历史上技术变革最剧烈的一个世纪。然而，最新的科技浪潮 —— 人工智能，对他们而言却常常遥不可及。

我坚信，我们最强大的科技能力并不仅仅是打造下一个独角兽企业，还是将强大的工具简化，让我们所爱的人也能使用。因此，我编写了这份AI使用指南，帮助 70 岁以上的长辈在手机上启用 AI 助手 —— 这个小帮手能让日常生活更轻松、更便捷，也更添乐趣。

若您认识可能从中受益的人，请将这份指南分享给他们。我们一起拥抱科技，让生活变得更加美好🌷

您的新 AI 助手：手机里的贴心帮手

亲爱的 70 岁以上的大宝贝们！您见证了一生的变迁 —— 如今，手机里又多了一位新帮手：AI 助手。

不妨把它想象成一位耐心友善的邻居，随时准备解答您的疑问、提供灵感，或陪您学习新事物。

别担心哦，不管您用的是啥牌子的手机，只要会点开微信，就能学会！来，跟着我一步步做，特别简单。

首先，咱们打开微信，看到最上面那个像小放大镜的按钮了吗？对啦，就是它，轻轻点一下。

点完之后会跳出一个长方形的小框框，点一下这个框框，看到里面有个小话筒了吗？长按它，咱们对着手机说两个字 ——‘元宝’，说完松开手就行啦，是不是很方便？

好啦，说完之后呀，手机会跳出一长串列表，咱们从上往下看找到”腾讯元宝”小程序，看到了吗？点一下它，咱们就来到腾讯自带的人工智能聊天界面啦，是不是很神奇？

第二步：开启您和AI的第一次对话

打开聊天界面后，底部会出现一个文本框 —— 就像给朋友发消息一样简单。

试试用元宝找菜谱，您可以这样问：

“已经立秋了，空气比较干燥，想做点润喉的菜。能给我 5 个顶级营养师推荐的菜谱吗？要带详细步骤的。”

几秒钟后，您就会收到贴心建议 —— 仿佛同时拥有了厨师、营养师和暖心朋友的陪伴。

第三步：用AI查询健康养生问题

您也可以用 AI 询问和健康养生相关的内容。比如：

“我是一位 80 岁的大叔，膝盖疼。能推荐一些来自权威医疗来源，比如三甲医院的安全居家锻炼方法吗？”

AI 会找到贴合您情况的、简单易懂的建议。

重要提示：AI 能提供不错的思路，但不能替代医生。采纳任何健康建议前，一定要咨询您的医护人员。

您的下一步行动

读到这里的您，不妨今天花 5 分钟，自己试一下，然后把这份指南分享给一位 70 岁以上的大宝 —— 父母、邻居或老友。陪他们一起设置好这个AI小助手。

这不仅仅是教他们使用一项技术 —— 更是给他们一份保持好奇、维系与时代的连接、收获快乐的工具。因为我们能分享的最珍贵礼物，不仅是知识，更是运用知识的信心。让我们一起拥抱科技，让生活更加美好🌷

视频版

70+ AI 使用手册链接地址

Our parents, grandparents, aunts, and uncles have witnessed the most remarkable century of technological change. Yet for many, the newest wave — artificial intelligence — still feels distant or intimidating.

Our greatest tech superpower isn’t just building unicorns; it’s making powerful tools so simple that someone we love can use them with ease. That’s why I created this guide to help anyone over 70 start using an AI assistant on their phone — a friendly helper that can make life easier, more connected, and more joyful.

If someone in your life could benefit, share this with them. Together, we can bridge the AI gap — one loved one at a time.

Your New AI Assistant: A Friendly Helper in Your Pocket

Hello to our young-at-heart readers over 70! You’ve witnessed a lifetime of change — and now, there’s a new kind of helping hand available on your phone: an AI assistant.

Think of it as a patient, friendly neighbour who’s always ready to answer your questions, give you ideas, or help you learn something new. It doesn’t matter if you have an iPhone, Samsung, or Google phone — this guide will walk you through, step by step.

Step 1: Getting Started with ChatGPT

We’ll use ChatGPT, one of the easiest AI assistants to try. You can set it up in two ways:

Method 1 — Using Your Internet Browser

1. Open your browser (Safari, Chrome, or Samsung Internet).
2. Tap the microphone icon in the search bar and say: “Chat G P T”.
3. Tap the link that says chat.openai.com.
4. Choose “Continue with Google” or “Continue with Apple” to sign up.

Method 2 — Downloading the App (Recommended)

1. Open your App Store (iPhone) or Google Play Store (Android).
2. Search for ChatGPT.
3. Look for the black-and-white spiral logo from OpenAI.
4. Tap “Get” or “Install” and sign up as above.

Tip: Once it’s installed, the hardest part is already done.

Step 2: Have Your First Conversation

When you open ChatGPT, you’ll see a text box at the bottom — just like sending a message to a friend.

Try this autumn example:

“Now that it’s autumn and the air is dry, I’d love to cook something good for my throat. Can you give me five recipes recommended by a top nutritionist, with detailed steps?”

In seconds, you’ll get thoughtful suggestions — like having a chef, a nutritionist, and a caring friend all in one.

Step 3: Use It for Health & Wellness Questions

You can also ask about gentle exercises, recipes, or hobbies. For example:

“I’m a man in my 80s with mild knee discomfort. Can you suggest safe at-home exercises from trusted medical sources?”

The AI will provide relevant, easy-to-follow advice tailored to your specific situation.

Important: AI can give great ideas, but it’s not a replacement for a doctor. Always consult your healthcare provider before acting on health advice.

Your Next Step

If you’ve read this far, take 5 minutes today to share this guide with someone over 70 — a parent, a neighbour, an old friend. Sit with them as they set it up.

This isn’t just about teaching technology — it’s about giving them a tool for curiosity, connection, and joy.

Because the greatest gift we can share is not just knowledge — it’s the confidence to use it.

Get senior-friendly tech help — sign up now!

Session Recap from the MANS International AI Application Series

WAIC 2025 Recap: AI Agents Take Center Stage

The 2025 World Artificial Intelligence Conference (WAIC), themed “Intelligent Era, Shared Future”, brought together over 800 companies and 3,000+ cutting-edge innovations. One of the biggest highlights? A rare public keynote in China by Geoffrey Hinton — Turing and Nobel laureate — who explored the provocative question: “Will digital intelligence replace biological intelligence?”

But the real spotlight this year was on AI Agents. Beyond the buzz, they represent a paradigm shift — from using AI as a tool to integrating it as a core operator in enterprise workflows.

Today’s session breaks down:

• What “AI-native thinking” really means
• The core logic of AI Agents
• How to evaluate and select the right AI Agent solution
• Real-world examples that show the path forward

I. From Digital 1.0 to AI-Native: A New Cognitive Framework for Business

Past: Digitization 1.0 — Humans Adapt to Systems

In the early digital era, humans were the glue between systems — manually logging into platforms, moving data, and pushing approvals. Think of a typical e-commerce operator juggling inventory, logistics, and finance portals daily.

Present: AI-Native — Systems Adapt to Goals

In the AI-native era, the AI Agent acts as the central brain. It understands goals (“launch a new product”), coordinates backend systems (SKU setup, inventory check, logistics alignment, budget approval), and only surfaces decisions that truly require human judgment.

Two Breakthroughs of AI-Native Thinking

1. Shift of Control: From AI as a tool to AI as the process orchestrator AI proactively breaks down objectives, allocates tasks, and drives execution. For example, a virtual assistant schedules meetings, books flights, and sends invites — all on its own.
2. Full-Chain Automation: Breaking data and workflow silos
   AI Agents bridge operations across departments — linking customer touchpoints, internal systems, and vendor pipelines for end-to-end automation. A customer order might automatically trigger inventory deduction, logistics scheduling, and post-sale follow-up.

Bottom Line: This unlocks smarter digitization, reduces operational load, and enables flexible, goal-driven workflows that adapt at speed.

II. Evolution of Human–AI Collaboration: From ChatGPT to Agents

There’s growing confusion around ChatGPT, Copilot, and AI Agents. But their differences matter — especially for businesses aiming to delegate work, not just generate content.

III. Enterprise-Grade AI Agents: Four Core Dimensions

To avoid hype traps and identify enterprise-ready AI Agents, tech leaders should evaluate solutions across four core dimensions: business logic depth, data security, deployment maturity, and risk governance.

1. Business logic penetration (Does it understand your domain?)
2. Data security & Compliance (Can you trust it with your core assets?)
3. Deployment readiness (Can you scale it without breaking things?)
4. Risk management (Will it break your systems or reputation?)

To access the full list of 12 evaluation criteria, key questions, and real-world examples for assessing enterprise AI agents, contact us directly.

IV. Real-World Case: AI Agents in Healthcare

Use Case 1: Clinical Diagnosis Support

Old Way: Radiologists manually review scans, taking 30+ minutes per report.

AI Agent Way:

• Instantly fetches data from EMR, PACS, LIS.
• Uses AI imaging models to detect and flag anomalies.
• Summarizes and presents only key decisions for doctor sign-off.
   → Result: 80% time saved, doctors become strategic decision-makers.

Use Case 2: AI-Accelerated Research

Old Way: Trial-and-error over a decade.

AI Agent Way:

• Mines millions of case records to identify drug candidates.
• Simulates drug–gene interactions.
• Generates real-time research reports and next-step suggestions.
   → Result: 40% faster from bench to market.

V. Conclusion: Becoming an AI-Native Company in 2025

Step 1: Understand the value.
AI Agents reduce cost, boost agility, and free humans for strategic work.

Step 2: Start small.
Test high-impact scenarios like customer support or supply chain first.

Step 3: Scale intentionally.
Expand across functions, building a digitally autonomous architecture.

The age of AI Agents is here. They won’t just support your business — they’ll help redesign it from the ground up.

Are you ready to lead the transformation?

2025 has been dubbed the Year One of AI Adoption. As artificial intelligence seeps into every corner of our work, learning, and daily life, education — humanity’s core tool for shaping cognition — now stands at a historic crossroads.

I. Kevin Kelly’s Forecast: Two Ways AI Will Reshape Education

Kevin Kelly, founding editor of Wired and bestselling author, believes the disruption AI brings to education will follow two main tracks:

1. Traditional One-Size-Fits-All Education Is Over

Students receive tailored AI support for weaker areas while diving deeper into subjects they’re passionate about. The purpose of education is shifting — from imparting standardized knowledge to nurturing individual potential: curiosity, creativity, collaboration, and systems thinking (the ability to “see both forests and trees”). In this model, education is not about memorizing content, but about training minds to think critically and act resourcefully.

2. The Reinvention of the University Model

As more highly educated individuals take on entry-level or even gig economy jobs (a rising number of master’s and PhD graduates are now delivery drivers), the value of a traditional college degree is being questioned.

Kelly argues that universities offer three main assets today: brand prestige, alumni networks and social capital. But in 25 years? He envisions:

• Skills and portfolios may outweigh diplomas. AI-powered lifelong mentors may replace professors for lifelong upskilling.
• Virtual platforms, the emerging “mirror worlds”, could replicate or surpass the social and learning experiences of physical campuses.
• Learning will become continuous, borderless, and embedded in our daily lives.

II. China’s Shift: From Degree Obsession to Skills-First Thinking

In 2025, Zhengzhou Railway Vocational and Technical College made headlines by admitting bachelor’s degree holders into its associate-level technical programs in high-speed rail maintenance. These programs lead to a lower-level diploma than the one students already have—but offer higher employment potential.

This counterintuitive trend challenges the long-standing belief that more education automatically equals better opportunities. The college first launched this program in 2022 and nearly doubled its enrollment in 2025, signalling a strong demand for skill-based education.

Last year, I visited Guizhou University of Applied Technology and was struck by the energy of its students and the relevance of its curriculum. For example, in the industrial robotics program, students must master both theoretical knowledge and hands-on skills in the installation, debugging, and maintenance of industrial robots. Graduates are in high demand. 

In the Yangtze River Delta region alone, over 6,000 companies currently use industrial robots. The talent gap for skilled technicians in smart manufacturing and high-end equipment is projected to widen dramatically over the next 3–5 years. In this context, hybrid talent—those with both domain expertise and technical know-how—are emerging as the winners of the job market.

III. The Rise of Vocational Education in the U.S.

This isn’t just a Chinese phenomenon. The U.S. is undergoing its vocational renaissance:

In the United States, a quiet revolution in career and technical education (CTE) has been unfolding for years. It gained national attention with the passage of the Strengthening Career and Technical Education for the 21st Century Act (commonly known as Perkins V) in 2018. The legislation allocates over $1.4 billion annually to support CTE and sends a clear message: a university is not the only path to success. Hands-on, practice-based learning is quickly becoming one of the most viable routes to upward mobility.

This policy shift is echoed in the job market. According to Fox Business, unemployment among recent U.S. college grads has risen to 5.8% — the highest since 2021 and significantly above the national average of 3.5–4%. Meanwhile, those with vocational associate degrees — especially in technical fields — enjoy much lower unemployment rates.

This demand surge has also fueled investor interest in “edtech for skills.” One standout example is Lumion, a company originally focused on student financing. Sensing the opportunity in vocational education, Lumion pivoted to provide SaaS solutions for technical schools. It raised $10.7 million in seed funding — backed in part by the state of Wyoming.

Since then, Lumion’s growth has been explosive: its revenue, customer base, and team have all tripled in the past year. Today, it serves over 260 schools across 29 industries and supports more than 100,000 students. Its software powers student recruitment, tuition management, and performance tracking. To date, it has facilitated over 140,000 tuition transactions, supported 1,690 school administrators, and is estimated to contribute $20 billion in workforce value. This thriving “policy + market + capital” ecosystem mirrors the momentum behind vocational education in China.

IV. Final Thoughts: From Degree Premium to Skills Premium

From Zhengzhou’s unconventional admissions, to Lumion’s VC-backed rise; from China’s high job placement for vocational grads, to America’s growing preference for skills over diplomas—a global shift is underway.

The common thread? Vocational and technical education is no longer a fallback option. It’s becoming a strategic choice.

This transformation is not just about education—it’s a collective human response to the AI-driven economic shift. In a world where algorithms reshape industries daily, skills that convert knowledge into action are proving to be more valuable than credentials alone.

Even more profound is how this evolution is redefining success itself. When practical skills outweigh university rankings, when professional dignity is no longer tied to academic pedigree, we may be witnessing the most meaningful form of educational equality: the ability for every worker to find purpose, value, and opportunity in an age of rapid change.

The future belongs to problem-solvers, not just degree-holders. Education, at its best, is not about producing diploma owners—it’s about cultivating people who can create real-world impact.

2025 年被标记为 AI 使用元年。当 AI 如空气般渗透进工作、学习与生活的每个缝隙，教育这一塑造人类认知的核心领域，正站在被重塑的十字路口。

著名科技畅销书作者、《连线》杂志前主编 Kevin Kelly 的前瞻性预判，正在中美两国的教育实践中逐渐显影 —— 一场从 “学历导向” 到 “技能为王” 的变革，已在全球范围内悄然拉开序幕。

一、AI 重构教育的未来图景：Kevin Kelly 的预言启示

在 Kevin Kelly 眼中，AI 对教育的颠覆将沿着两条主线展开。

其一，是个性化教育的全面普及。未来课堂将彻底告别 “统一进度” 的标准化模式，学生依托 AI 按照自身节奏推进学习 —— 基础薄弱处有针对性辅导，兴趣擅长点能深入拓展。这意味着教育核心目标将从灌输共性知识，转向强化每个人的独特性：呵护好奇心、激发创造力、锻造协作能力，培养 “见树又见林” 的全局视野与 “见终局” 的前瞻思维。

其二，是大学形态的深度重塑。当 “高学历者从事基础职业” 的现象渐增（“研究生、博士生加入外卖队伍” ），“上大学的意义” 被重新追问。Kevin Kelly 将当下大学价值归结为三点：名校光环与校友网络、社交圈子构建、非线性且高度虚拟化的学习体验。

而未来，学历价值可能持续弱化 —— 一个人参与过的活动、展现的能力特质，将比文凭更有说服力。他甚至畅想，25 年后社交与学习可能分离：AI 专属助理成为 “终身导师”，在线课程完成知识传递，“镜像世界” 等技术创造超越校园的社交体验。未来终身学习将成为常态，教育不再是 “青春限定”，而是与生命同频的持续成长。

二、中国教育新趋势：从学位竞争到技能为王

Kevin Kelly 的预言，正在中国教育场域中上演具象化实践。

2025 年，郑州铁路职业技术学院的招生简章意外刷屏 —— 该校首次面向本科毕业生开放 “动车组维修技术”“高速铁路综合维修技术” 等专业，毕业后颁发专科文凭。这一 “本科进、专科出” 的选择，打破了 “学历越高越好” 的固有认知。事实上，该校自 2022 年起试点本科招生，2025 年计划招生 135 人，较 2024 年增长近一倍，释放出明确信号：技能型人才培养正在抢占高等教育 “C 位”。

这并非个例。去年感谢古姐，我有幸跟随贵州电视台前往贵州应用技术学院参加活动。校园规模宏大，环境宜人，学生们充满活力。拿工业机器人应用和维护专业举例。该专业学生不仅要掌握专业理论知识，还要精通现代工业机器人安装、调试、维护等实操技能，这类毕业生在市场上供不应求。单在长三角地区，使用工业机器人的企业就超过六千家，预计未来3-5年该领域高技能人才缺口将持续扩大。

在智能制造、高端装备等战略新兴领域，”专业知识+实操技能”的复合型人才将成为最大赢家。

三、从中国到美国：职业教育的全球共振

这种技能导向的教育变革并非中国独有。在美国，一场由政策、市场与资本共同推动的职业教育转型早已启动 ——2018 年《加强 21 世纪职业技术教育法案》（PerkinsV）的通过，标志着联邦政府将职业教育（CTE）提升至战略高度：该法案每年投入超 14 亿美元，明确传递出 “大学并非唯一成功路径” 的信号，而以实践为核心的技能培养，正成为对接现代职场的重要通道。

政策风向的转变，在劳动力市场数据中得到直接印证。据福克斯商业新闻报道，美国应届大学毕业生失业率近期已升至 5.8%，为 2021 年以来最高，显著高于该国 3.5%-4% 的整体失业率；反观拥有职业副学士学位的群体，尤其是掌握特定技术的毕业生，失业率远低于普通学位持有者。

市场需求的爆发，更催生了职业教育科技赛道的资本热潮。2020 年成立的 Lumion 就是典型案例：这家原本聚焦学生融资的平台，敏锐捕捉到职业教育的增长红利，果断转型为职业技术学校提供 SaaS 解决方案，并在种子轮获得 1070 万美元融资（投资方包括怀俄明州政府）。

转型后的 Lumion 增速惊人：过去一年收入、客户群和员工数量均增长两倍，目前已服务 29 个行业的 260 多所学校、10 万余名学生。其核心产品覆盖招生（全流程数字化管理）、付款（多元支付方案提升入学率）、学生信息系统（全周期数据追踪）三大环节，累计处理 14 万笔学费支付，支撑 1690 名学校管理者，预计带来 200 亿美元人力资本价值。

这种 “政策输血 + 市场造血 + 资本活血” 的生态，与中国高职教育的扩张形成跨洋共振。

四、结语：技能溢价时代，教育回归价值本质

从郑州铁路职业技术学院的招生扩容，到 Lumion 的资本追捧；从中国高职毕业生的就业优势，到美国职业教育群体的职场竞争力 —— 中美两国虽教育体系各异，但在 AI 重构全球劳动力需求的浪潮中，都指向了同一个趋势：职业技术教育正从 “备选方案” 升级为 “战略选择”。

这场变革的深层意义在于：它正在解构工业时代建立的学历崇拜体系。当算法和自动化不断重新定义工作岗位，教育回报率的衡量标准正从”学历溢价”转向”技能溢价”，新的人才价值公式愈发清晰：解决问题的能力＞理论知识的储备。

教育的终极命题或许正在回归本质：不是制造文凭持有者，而是培养能创造真实价值的问题解决者。当职业尊严不再取决于学位等级，当个人价值更多由实际贡献而非学历背书决定，这或许才是教育公平最深刻的体现——让每个人都能在技术变革中找到自己的价值锚点。

If you’re investing in AI, building with it, or worried about being replaced by it, there’s one principle you need to understand: Verifier’s Law.

Last Friday, we hosted a session as part of the MANS International AI Strategy Series, where we unpacked Verifier’s Law — developed by Jason Wei, a leading AI researcher who recently left OpenAI to join Meta’s Superalignment team.

This under-the-radar principle explains why AI is transforming some industries faster than others — and how to anticipate what’s coming next.

What is Verifier’s Law?

“Any task that is possible to solve and easy to verify will be solved by AI.”
— Jason Wei

In simpler terms: if a machine can be trained to judge whether an output is good or bad—quickly, accurately, and at scale—it will master that task faster than we expect.

What Is the “Asymmetry of Verification”?

The Asymmetry of Verification refers to tasks that are difficult to perform but easy to verify. In other words, while producing the solution requires significant time or expertise, checking whether the solution is correct is fast and straightforward.

Here are some everyday examples:

• Sudoku: Solving a puzzle can take 20 minutes, but verifying a completed solution takes just 2 seconds.
• Software development: Writing backend code may take weeks; running it to check if it works takes only seconds.
• Research questions: Finding a reliable answer might take hours; verifying it can take just a few clicks.

This asymmetry is exactly what today’s most advanced AI systems — like ChatGPT, GitHub Copilot, and AlphaEvolve — are harnessing at scale. They don’t just solve problems — they thrive in domains where verification is easy, allowing them to learn faster, adapt quicker, and outperform expectations.

Five Criteria That Make a Task “AI-Solvable”

According to Verifier’s Law, developed by AI researcher Jason Wei, certain tasks are particularly well-suited for AI. They share five key characteristics:

1. Objective truth — There’s a clear standard for what’s correct, with little room for disagreement.
2. Fast verification — You can quickly determine whether a solution is right or wrong.
3. Scalable verification — It’s easy to verify many outputs at once, enabling rapid learning.
4. Low noise — The verification process reliably reflects the true quality of the solution.
5. Continuous reward — Solutions can be ranked along a spectrum from poor to excellent, not just right or wrong.

If a task meets three or more of these criteria, it’s a strong candidate for AI automation — and likely to be transformed sooner than expected.

Case Studies: How OpenAI and Google Brain Apply Verifier’s Law

AlphaEvolve (Google Brain)

Challenge: “What’s the smallest hexagon that can fit 11 unit hexagons?”
 Insight: Solving this geometric optimization problem is difficult — but verifying a solution is instant.
 Result: AI explored solutions at scale, uncovering designs that human experts hadn’t imagined.

GitHub Copilot (OpenAI)

Challenge: Automating code generation
 Process: Copilot generates code → automated tests verify correctness → feedback refines future outputs.
 Result: A continuous improvement loop powered by fast, scalable verification.

These breakthroughs didn’t just rely on AI’s intelligence — they were engineered around verifiability, accelerating progress in ways traditional approaches couldn’t.

What This Means for You

For Investors

Use Verifier’s Law as a lens for evaluating AI startups. Ask yourself:

• Can the AI’s output be tested and measured quickly?
• Are feedback loops built into the product design?
• Does the task satisfy three or more of the five AI criteria?

Invest where verification is fast and iteration is cheap. That’s where AI compounds value quickly.

For Founders

Want to build a high-impact, defensible AI product?
 Focus on:

• Tasks that are time-consuming or expensive to solve manually
• But cheap and fast to verify

Don’t just solve problems — engineer them to be verifiable.
 That’s where you unlock scalable feedback and competitive advantage.

For Professionals

Concerned about being replaced by AI? You’re not alone.
 Here’s the hard truth:

AI will dominate tasks where solutions are easy to verify.

Resilient jobs will involve:

• Ambiguity, complexity, or nuance in defining a “good” result
• High-context decision-making
• Human trust, ethics, or emotional intelligence

The more difficult it is to verify your work, the longer it will take for AI to replace it.

Ready to Apply Verifier’s Law?

We’re opening 2 spots for investors who want expert support evaluating their AI portfolios — and 3 spots for founders seeking feedback on whether their project has real market potential.

Get personalized insights. Build smarter. Invest wiser.
 → Message us to claim your spot.

The MANS International AI Strategy Series is currently available exclusively to our valued clients, strategic partners, and registered members.