Why China's 2.1 million open source contributors still don't own the top frameworks

Why almost all famous open source projects are written by foreigners

I hardly dare to broach this topic; the moment I do, I get attacked or subtly mocked.

One camp says 'we simply lack an open source spirit,' another says 'they have money and free time.'

I've long subscribed to these two views myself, but the data now says otherwise.

Let's put the data on the table

According to the 2025 China Open Source Development Deep Report, China has more than 2.1 million active GitHub developers, ranking third globally. (Image: source from internet; remove if infringing.)

Note this is only GitHub's data, not counting domestic hosting platforms. Including Gitee, AtomGit, GitCode and others, the total is estimated to exceed 3.5 million, ranking second globally. Active open source projects number roughly 4.25 million, up 40.58% year-over-year, contributing 12.38% of global activity.

So we're not absent; we participate quite actively. When AI summarized this data for me, it produced a line I find spot-on: many contributors, few owners.

That is, we excel at submitting PRs, fixing bugs, and writing docs in others' repos, but rarely do we open our own repo, define an API, and nurture a global developer base. Scrolling the top of the star rankings for frameworks initiated by Chinese teams feels like hunting for salad in a hotpot restaurant. They exist, but they're not on the front page.

Frameworks aren't built; they grow

Many imagine a framework as the product of a genius's midnight epiphany: on some date, a great mind wakes from a dream where a white-bearded immortal (or a crucifix-wielding one) commands, 'Invent such-and-such framework.' But obviously, most famous open source projects didn't arise that way. Nearly every top-tier framework grows on top of a particular platform layer.

• React was written by Facebook to solve its own page performance issues.
• Kubernetes abstracts the lessons from Google's internal Borg system.
• Spring came from Rod Johnson's book and the code that accompanied it.
• .NET is simply part of Microsoft's own platform.

So the pattern is clear: whoever defines the runtime defines what a framework can look like. For the past two decades, the power to set standards for the foundational layers—programming languages, browser engines, cloud platforms, package managers—has largely slipped from our hands. So the outcome is natural: we're absurdly strong at the application layer but weak at the framework layer. This is not a capability problem; it's a positioning problem. Someone standing on the fourth floor can hardly casually re-pour the ground floor's foundation.

Talent isn't what really holds people back

Silicon Valley has perfected a smooth loop: open source → personal reputation → talks, job hopping, funding, sponsorship → full-time maintenance → better project → more reputation. Evan You used Patreon sponsorship to maintain Vue full-time, and has since founded a company. Vercel, Supabase, and HashiCorp treat open source as a customer acquisition cost. In their context, open source is an asset that can be converted into money.

But in China, this chain is broken—broken in three places:

1. No star column on performance review sheets. Your performance is measured by requirement delivery rate, launch quality, and downtime length. A framework written over three months carries roughly the same weight on a performance review as fixing the boss's printer or watering his plants.
2. The company holds veto power. Legal worries about non-compete and compliance, security teams worry about audits, and the boss worries about competitors using it commercially. So most projects incubated inside a company never reach release.
3. A cycle forms. Open source → conference marketing → abandoned after six months. The more such projects, the less global developers trust our projects, making it harder for the next batch to gain traction.

More critically, you won't get a promotion for building a wheel. But you might get called in for a wheel-related incident, or blamed by colleagues. That's why so many people's dream frameworks end up stuck in local branches.

Writing code isn't the same as nurturing a community

One step is gravely underestimated: distribution. A project's ceiling is often set not by code quality but by community operations. Is the English README clear? Do issues get replies within 24 hours? Is there a roadmap and RFC process? Will people on Hacker News and Reddit help spread the word? These are dedicated jobs that require investment.

Almost no company in China has such a role. Many technically strong projects lose international developers the moment they click in, see documentation only in Chinese, and politely close the tab. Don't think foreigners are arrogant. GitHub is an English-language site, and there are more English-speaking developers worldwide than Chinese-speaking ones. If the technical gap is only 20%, the distribution gap is often 200%.

Money-and-leisure and involution are each half right

Now about 'money and leisure'—this framing actually reverses cause and effect. Most framework authors aren't independent; they're employed by companies to write them. React is Facebook's salary, Kubernetes is Google's salary, .NET is Microsoft's salary. The reclusive genius you imagine is essentially a salaried contractor. A few individual heroes do live off sponsorship, but that's a result of success, not the cause.

As for 'involution' (neijuan), that's real, but its damage is often misunderstood. Two hours a day indeed can't write React. But more fatal than time is organizational disincentive and systemic non-reward. You don't lack ideas; you lack an outlet for them. So money and leisure might be one factor, but certainly not the main one.

The lane-change period has already begun

The good news is that the landscape might be reshaped. In 2025, six of the ten fastest-growing open source repositories in the world are AI infrastructure. And in this layer, China is already a major player! Qwen, DeepSeek, Dify, RAGFlow, PaddlePaddle, plus the large number of Chinese contributors in the vLLM and llama.cpp ecosystems.

Why can this layer rise? Because the platform is new, not yet locked by two decades of historical ownership. This is the first genuine lane-change opportunity in twenty years. Moreover, with everyone focused on AI and leading companies racing to open source, the value of keeping a trained model closed source has dwindled. When open source generates momentum, it can feed back to the company, gradually shifting executives' mindsets.

Conclusion

In truth, most companies today aren't very interested in open source, and even the open source projects that do exist receive little care. Widely used projects routinely stop updating overnight, and issues go unanswered for eons. Perhaps we should support a whole open source ecosystem that closes the business loop, rather than endlessly blaming domestic developers for lacking open source spirit.

So I suggest that anyone going into open source must focus on two things: First, don't ask 'Is anyone in China doing this?' Ask 'Is this a pain point for developers worldwide?' Vue started as a side project Evan You wrote while at Google, with an unremarkable beginning. Second, treat English documentation and issue responses as top priority. These two things aren't 'geeky' or cool, but they determine whether your project is used by ten people or a hundred thousand.

And one more thing that may sting: don't wait for company approval. Most breakout projects in China are built first, then force the organization to acknowledge them.

I'm Li Jianyi, a front-end developer who's been coding for ten years and still feels like a failure. Follow my WeChat public account of the same name.