360 Group’s Zhou Hongyi Says He’s “Coding with AI” for Hours — The OpenClaw Era and the Rise of Atomic Skill Packages

The visit and the message

It has been reported that a delegation of more than 100 prominent entrepreneurs from a Yale scholars-and-alumni group recently visited 360 Group (360集团), where founder Zhou Hongyi delivered a candid, wide-ranging talk on what he called the OpenClaw (龙虾) era of software evolution. Zhou said he now spends whole days “coding with AI,” sometimes running for ten-odd hours, driving intelligent agents from his phone to develop software. “When it’s smart, it can generate thousands of lines in an instant, sometimes surpassing me,” he said, “but when it makes mistakes it’s maddening — it will randomly delete code. What takes it five minutes, I may have to review for an hour.”

Zhou reportedly told the visitors that many Silicon Valley CEOs and investors are similarly rolling up their sleeves and using AI to write code, sometimes orchestrating dozens of agents concurrently. Why are senior leaders doing the hands-on work themselves? Zhou’s reasoning was blunt: without deep, personal use of AI tools, managers cannot grasp where this technology will steer their industries.

What this means for engineering and management

Zhou outlined a near-future software model broken into “atomic” skill packages that autonomous agents will call on as needed. When current toolchains fall short, agents may even write new code autonomously. The implication is a shift in engineers’ core tasks from routine implementation to orchestration — designing, composing and supervising fleets of intelligent agents — and a new demand for what Zhou called “silicon-based leadership” (硅基领导力).

This vision has practical and geopolitical resonance. As companies in China accelerate internal AI adoption, managers’ firsthand familiarity with agentic workflows becomes a competitive advantage. At the same time, amid ongoing U.S.-China tech tensions and export controls on advanced chips and components, rapidly fielding software-centric capabilities could be a hedge against hardware constraints. Reportedly, the era of tool-first leaders is arriving — but it will bring fresh operational, security and governance questions as organizations scale agent-driven development.