“All-in-AI” and 知行合一

I believe AI is one of those technologies that comes along perhaps once or twice in a century and changes how society works. As a student of organization theory, I expect that change to be organizational as much as technical. AI makes some kinds of intellectual work, once dependent on scarce human expertise, much more widely available and easier to scale. I can give more work to AI as my budget and access allow; I cannot expand my own attention, energy, or capacity for judgment at the same rate. I have to decide how to divide the work, what workflows and routines to build, and where to spend my own brainpower. Questions usually associated with managers and team leaders now arise even for someone working alone.

People are already using AI to make films screened during the 2026 Cannes festival,¹ run a business with a single employee and a reported 10,000 paying customers,² create explorable 3D neighborhoods,³ generate interactive livestreams using MiniMax-based models,⁴ and support electricity-grid dispatch in Shenzhen.⁵ These uses span different technologies and settings. As a researcher and teacher, I cannot take for granted that my own understanding of AI is keeping up with what people are already doing with it.

If I claim to research AI and contribute new knowledge about it, but my experience ends with chatting to ChatGPT, using 豆包 (Doubao) casually, or asking Claude Desktop to write code, I would worry about the distance between my research and the phenomenon I claim to understand. I take inspiration from 知行合一, the unity of knowing and acting associated with the Ming-dynasty philosopher Wang Yangming.⁶ In my own scholarly practice, that means working with frontier capabilities myself and letting that experience inform the questions I ask. I want to understand how people decide what to delegate as models improve, how they judge work they could not have produced alone, and how learning from one project becomes knowledge that agents can use in the next.

For teaching, when students use frontier models every day, their experience shapes the questions they bring into the classroom. If a student shows me something they have built with AI, I want to understand how they made it, examine the choices behind it, and help them take it further. My knowledge of theory and methods gives me something to contribute, but I need practical experience with AI to connect that knowledge to the work in front of us.

That is why I decided to go all-in on AI. I am now proud to call myself a builder. My focus is on how these incredible, sometimes intimidating frontier models can help us become better social scientists and better teachers. There is a selfish side to this, too: I want to equip myself with the capabilities to take part in whatever comes next. For me, being a lifelong learner means attempting things, finding out where my understanding falls short, and returning to the work with what I have learned. Building my own AI-native conglomerate is how I have chosen to do that.

Sources

The Festival de Cannes confirms both the official Special Screening and use of generative AI in Steven Soderbergh’s John Lennon: The Last Interview (15 May 2026). Nino Leitner documents the industry screenings of Higgsfield’s Hell Grind in CineD (28 May 2026); that film was outside the Festival de Cannes official program. ↩
Polsia is the example here. In a 2026 founder update, Ben Cera (also known as Ben Broca) reports reaching 10,000 paying customers. His GTMnow interview with Sophie Buonassisi (1 June 2026) describes him as the only employee, with AI handling coding, customer support, email, and other operations. He also describes help from outside infrastructure providers; “single employee” does not mean no external human support. The customer count is founder-reported. ↩
Matt Shumer (7 September 2026), “How to Build 3D Worlds with Astra.” His creator account describes a walkable reconstruction of his childhood neighborhood and a 3D Manhattan, using Blender for assets and Three.js or Unreal Engine for scene assembly. ↩
fal, “H3 Max Director: Realtime Video You Direct as It Generates.” The provider describes fal.live as an experimental AI livestream with viewer-directed channels. It uses fal’s H3 Max Director, derived from the MiniMax H3 model family. The livestream project is operated by fal. ↩
Shenzhen Municipal Government portal (27 May 2026), “深圳打造‘人工智能+’能源发展新高地.” The report identifies Tianxuan–Lingxi, a model developed for urban electricity-grid dispatch by China Southern Power Grid’s Shenzhen Power Supply Bureau. ↩
Bryan W. Van Norden, “Wang Yangming,” §3, “Unity of Knowing and Acting”, Stanford Encyclopedia of Philosophy. Wang’s doctrine concerns ethical cultivation; its application here to my own scholarly practice is a personal interpretation. ↩