How “Lobster” and vibe coding are rewiring how people build software — from novices to entrepreneurs to engineers

An unexpected democratization of building

Huxiu (虎嗅) ran a wide-ranging conversation with the WeChat account Shengdong Huopo (声动活泼) that captures a striking shift: tools like OpenClaw (commonly nicknamed “小龙虾” or “lobster”) plus the rise of so‑called “vibe coding” are lowering the barrier to creating useful software. Non‑technical staff at a small media company reportedly used AI to produce working prototypes after a company hackathon — audio editors, brainstorm tools, visual assets — all built by people who previously wouldn’t touch code. The result: more people can make small, bespoke tools that solve immediate workplace pain points. Who needs a full engineering sprint for every idea anymore?

OpenClaw, agents and the new software logic

OpenClaw (小龙虾) is being talked about as more than a clever toy. It is open‑source and, it has been reported that its community contributions have exploded — reportedly thousands of developers are pushing commits — and that scale matters. Unlike classic apps that wait for user clicks, OpenClaw’s architecture emphasizes timers, long memory and “skills” — small, repeatable procedures the agent can teach itself or be taught. That is the essence of what Andrej Karpathy has dubbed “vibe coding”: programming by intention and iteration, guided by AI rather than line‑by‑line typing. Is this the end of programmers? Not yet. It is, however, a fundamental rethink of how software can be composed and evolved.

Different users, different gains — and limits

For product people and founders, vibe coding acts like a rapid prototyping therapist: it helps crystallize messy ideas into working demos and clarifies requirements through interaction. For engineers, the new tools accelerate routine work but still leave the hard problems — scalability, correctness, maintainability — in their hands. The Shengdong Huopo team’s own experience building an automated news‑scraping and idea‑recommendation pipeline shows both sides: a fast, fruitful prototyping loop paired with the recognition that long‑term, reliable systems still require engineering rigor. And models themselves remain imperfect — repetition, hallucination and fragility are common — so OpenClaw today is a hinge moment, not the finished product.

Bigger picture: ecosystems, geopolitics and the future of creative work

This shift is playing out in China’s broader AI ecosystem as companies and developers lean into agent‑style services and open‑source frameworks. Against a backdrop of intensified US‑China tech competition and export controls, open projects offer a path to local experimentation and rapid iteration. But human judgment still matters: the non‑quantifiable instincts of editors, designers and senior creators are hard to reduce to skills or memories. The real question now is not whether AI will code, but how roles will be reallocated — and whether societies can capture the productivity gains while preserving the elements of work that remain irreducibly human.