"AGI Is Here? I Used It for a Week and My Scalp Went Numb" | A Conversation with Haoran Zhang: Cofounder of Moxt

🚥 This episode of Crossing features Haoran Zhang, co-founder of Moxt.

Last week, while showing me Moxt's beta, Haoran said something that stunned me: "AGI is already here. People just haven't found the right way to access it."

This comment came after his team built Moxt in three weeks post-Lunar New Year. Haoran described himself as "having my mind blown after using Moxt for a week." He also mentioned encountering "fanatical Moxt users whose eyes literally light up" — something he never experienced with his previous products, Motiff and Paraflow.

We've used ChatGPT and Manus mostly as individuals. But if entire teams and organizations are going to use AI together, we'll need fundamentally new software. Moxt's AI Native Workspace is one attempt at this.

What exactly is Moxt — and what isn't it? Why would someone who's spent years building products feel like "the future has already arrived"? And how does the new way of working it represents fundamentally differ from the Lark, Notion, ChatGPT, Manus, OpenClaw, and Claude Code we're familiar with?

If you're working in AI startups or investing in them, or if you want to understand how AI-native teams will work and what new AI tools they'll adopt, I believe this episode will offer something valuable.

🎬 Our video podcast is now live on Koji Yang Yuancheng's channels on WeChat Video, Xiaohongshu, Bilibili, YouTube, and other platforms.

📒 The transcript will be published on the Crossing WeChat official account.

🟢 00:00:32 Rapid Fire

Age, alma mater, MBTI and zodiac sign, one-sentence pitch for Moxt, revenue and order volume, team size, pre-founding experience

🟢 00:02:22 AGI Is Here, Just With the Wrong Interface

"My mind was blown." What does "AGI is here" mean? "In 90% of industries, AI can handle 90% of 90% of people's work" — by this standard, his day-to-day experience is: it's genuinely arrived. Why did using Moxt for a week blow his mind? Not just because it can do more, but because it makes you question at every moment of work: why am I still working this way?

This shock is the main reason Haoran has been in what he calls a "manic state" lately.

🟢 00:04:17 An AI Colleague Appears in My Meeting for the First Time

The old collaboration pattern: send documents, write comments, then find you to talk — this workflow has nearly disappeared from their team. Their one-on-ones: AI drafts the document, two people talk, AI records the whole conversation, and by the end the document has already updated itself automatically. "Far beyond what I previously thought possible."

One detail: when he talks to the AI, he no longer needs to explain "what is Moxt" or "who is Koji" — the AI looks it up itself. No need to feed it background.

This reminds him of when he first learned about AI and had to write "you are a professional..." in prompts — that approach is already obsolete today.

🟢 00:08:30 What Is Moxt?

In one sentence: an AI-native workspace where you can assemble your AI team.

In Moxt, everyone has a贴身 assistant that must be named "MOMO," which knows all your information. Other AI colleagues can be created by different people, appear in Slack channels, and proactively DM you.

What's the most fundamental difference from Notion and Lark?

🟢 00:11:03 "Fundamentalism"

A Word document isn't 500 words to AI — it might be 50,000 characters. The actually useful information is drowned out. That's information loss.

Moxt's "fundamental" rule one: documents use only Markdown, tables use CSV, visualization uses HTML. "Markdown/CSS/HTML are the new Word/Excel/PPT."

Fundamental rule two: use a file system (OS-level), not Notion-style knowledge base structures — to AI, Notion's hierarchical structure is a labyrinth.

Feed these two rules to AI, and "magic emerges naturally" — it's not that intelligence is insufficient, but often that the context is fed in the wrong format.

🟢 00:16:09 The Tools Moxt Has Killed

They no longer hold team stand-ups — because Moxt knows what everyone is doing every day, there's no need to "resync." They built their own internal dashboard and completely abandoned Jira after ten years of use — and it only took three days.

A more radical question: why even use a dashboard? Dashboards exist to solve the information sync problem that stand-ups address. But if AI already knows all the context, does information sync still need a dashboard as a "container"?

🟢 00:28:08 Haoran's Five AI Colleagues

Run Manager (fierce kitten avatar): helps him sync management information, proactively nudges on project progress. Some colleagues have started not coming to him directly, but communicating with this AI manager instead.

Growth Heavenly King (gold medal salesperson): goal is one sentence — "find 1,000 paying users for Moxt, make a plan." It even built its own CRM table, turning it from Excel into an HTML dashboard.

Deep Thinker: reports every two or three days, synthesizing internal OKR changes and external competitive dynamics, making PPTs page by page for him.

Miss Creative: dedicated to divergent thinking, housed in a separate space — the wilder the better.

The Critic: examines with a harsh eye whether Haoran himself is working on the most important things each day. The biggest difference from prompting GPT "you are a critical thinking expert" is that it lives in this space and has real-time access to all context.

🟢 00:37:06 How Moxt Was Born

The starting point was a strange phenomenon: after the team started using Claude Code and Cursor, Slack channels began filling up with people sending each other markdown files — having finally entered the era of collaborative work, everyone was back to sending raw files.

Version one was extremely simple: build a markdown-based cloud collaboration system so people would stop passing markdown files around. That was it.

One and a half engineers, one proposal, originally just solving an internal problem — the product only launched publicly three weeks ago.

🟢 00:46:02 Why Is Moxt's Document Editing So "Hard to Use"?

Many early users complained: your documents require clicking top-right to edit, then clicking save, way worse than Notion.

But this is deliberate design: when AI context is sufficiently rich, manual editing becomes an extremely low-frequency operation. Having AI write directly is far more efficient than editing yourself.

Their interaction logic: like an emperor reviewing memorials, you underline and comment in the document, AI colleagues read the comments and revise directly — even two colleagues' MOMOs can debate each other in the comment section.

Where does the brand name Moxt come from? More Conxt — mo and xt, hoping users will remember them more easily.

🟢 00:50:42 The Disappearing Ways of Working

Team stand-ups gone, Jira gone, manual documents dwindling, IM also experiencing friction... so what ultimately remains?

Haoran asked his own MOMO: "When humans barely execute anything anymore, what do we have left?"

The white-collar day, two or three years from now.

Sudden insights and aesthetic judgments often emerge from aimless conversation — this反而 becomes more important.

🟢 01:00:46 The Biggest Enemy Is Time, Yet He Hopes People Move Slower

He wants more people to start using Moxt quickly, but simultaneously hopes people discover this more slowly — because the faster big players take notice, the more boring the battlefield becomes.

A month ago, if someone had told him the Moxt story, his first reaction would have been: "How is this different from Lark plus Lobster?" — so he still needs to spend considerable effort explaining this today.

On Lark, Notion, and other major players: he doesn't deny they'll see good AI-driven growth; but he sees this as layering AI onto old paradigms, not rebuilding from an AI-native starting point.

On big tech competition: big tech can come, but will they send their best people? They fought big tech when doing education too, "and it wasn't that scary."

🟢 01:04:43 AI Amplifies People, or Replaces Them — This Is Moxt's Bottom Line

He wrote one sentence in a document as "constitution": at no point should we create content suggesting "AI can replace humans"; your better role is to amplify people.

He saw some products' Pricing pages feature a comparison table of AI employees vs. human employees — he doesn't want this to happen with Moxt.

A metaphor for the future: going to a pottery studio today is paid leisure, but once upon a time, pottery-making was highly productive labor. Will programming become the pottery-making of 30 years from now?

---

Welcome to subscribe to Crossing: 🚦 We follow the industry transformations and new entrepreneurial opportunities brought by the new wave of AI technology.

🚦 Crossing is Steve Jobs' metaphor for Apple — standing at the intersection of technology and liberal arts, where great products are born. AI is transforming every industry. We seek out, interview, and bring together a new generation of AI entrepreneurs and active participants in the AI era. Together with them, we explore and embrace the new changes, the new possibilities.

👦🏻 Host Koji: I founded Crossing and launched AI Hacker House, a community space for a new generation of AI entrepreneurs. I serve as Venture Partner at ZhenFund. I believe technology, especially AI, represents the greatest value creation opportunity of our generation. Koji's Jike, Koji's website

👧🏻 Host Ronghui: I co-founded Crossing. I've worked at a USD VC and spent five years as a Silicon Valley correspondent, following tech development and business stories. Welcome to chat with me and exchange ideas. Ronghui's Jike