KubeCon + CloudNativeCon China – does AI threaten the open source movement? Exclusive Q&A with CNCF, Huawei, and Solo.io

Just by The Peak on Hong Kong island – jump in the tram to the summit, where the views are breathtaking – you will find the Hopewell Hotel. A well-named venue for the Cloud Native Computing Foundation's KubeCon + CloudNativeCon China event, as the CNCF celebrates its tenth year and looks towards the future.

Away from the conference floor, buzzing with excited coders, developers, and engineers – most of them Chinese, but with a handful from Europe, India, North America, and Africa – I pulled up a chair to hear from three of the event's hosts and keynote speakers.

This was CNCF CTO Chris Aniszczyk; Kevin Wang, Lead of the Cloud Native Open Source Team for Shenzhen-based electronics giant, Huawei; and Lin Sun, Head of Open Source for cloud-native API management platform, Solo.io.

AI coding and the future of open source expertise

My first question had been brewing in my mind throughout the conference. First of all, CNCF research (see my previous report) finds that many open source users are experienced European coders, with 10+ years in the industry under their belts. But all the buzz from the AI sector today is about vibe coding and automation, with CEOs claiming that AI will be able to do 100% of coders' jobs by the end of this year.

Meanwhile, some companies claim they no longer employ juniors, and already default to AI, while some coder blogs report expertise ebbing out of the industry as those juniors who do find jobs seemingly lack the skills to check the AI's work, explain its decisions, or propose alternative solutions. A perfect storm of challenges?

So, how do all these megatrends affect the world in which the CNCF operates? And Kubernetes users, whose lifeblood is expertise and navigating complexity? Put simply, does AI-first, or AI-only, coding threaten the very culture of the open source movement?

Aniszczyk says:

We have our own maintainers who use a variety of different tools, including AI and code-assisting tools. I don't know if that falls into the 'vibe coding' camp, but a lot of developers are using these tools within our community. If it makes their job easier, if it helps in some way, then great.

But in CNCF, we view ourselves as providing the building blocks. So, the people out there that are building vibe coding tools… you know what? What people build with our projects, we don't necessarily care. But we'll use some of that output to make our lives easier.

I think you're maybe asking a broader question. Like, how does open source get impacted by more stuff being produced? It just means there is more code out there. There's already a tonne of code. And people will just use what's best.

Maybe – or maybe not. But there is a cultural dimension too, he adds:

The CEO of Shopify, Toby [Tobias Lütke] recently stated that everyone in that company – engineering or not – has to be AI-first, or AI native, and has to learn how to use these tools and get value out of them, otherwise they're going to be left behind or no longer be an employee.

Companies who don't have that approach are going to fall behind because they're going to ship software. […] Folks that are not AI native are going to lose to their competitors.

Huawei's Wang says:

We, a lot of the engineers, have been discussing that language is just a kind of tool, especially when you need to switch to another development language. Previously, it took time to learn well, especially if you don't know some of the functions or methods provided by that language. But with AI as an assistant, it helps you quickly have an easy prototype to start.

But in the end, you still need to do the design, and that's very important. For most open source projects, we care a lot about open architecture, and also need to implement and compare multiple implementations and different requirements.

Sometimes you need to make a choice, not just a way to implement something. And I think that part is becoming more and more important.

Of AI assistants' role, Aniszczyk adds:

None of these tools is perfect yet, by any means. They do help, but they're not going to help with maintenance or, potentially, with security issues. Or with the day-to-day job of what a lot of our maintainers deal with. The production and maintenance of projects is a very different task.

Fair enough. But my underlying point is this: the leaders of the AI industry – which the CNCF has been talking up throughout the conference – are claiming that AI will soon be able to do 100% of coding work, as soon as this Christmas.

Granted, those CEOs are salesmen, and they have investors desperate to demonstrate ROI. But all this must be having an impact on the culture of the open source movement, at least, and the passion that people have for their work? So, are they the old guard now, not the new?

He responds:

It affects [the industry] in that it is more helping hands. Whether they're humans or agents or bots, that's a good thing, right?

Maintainers have always had to deal with contributions from all over the world and from all different skill sets. Sometimes you get a drive-by contribution, or someone fixes a spelling mistake, or someone implements something that you don't really want, and it potentially ruins the architecture. Things like that have always existed. It is more that these tools accelerate the burden on maintainers to manage and deal with it.

People aren't necessarily firing engineers or, honestly, replacing them with agents that are doing all the work [this contradicts what some business leaders are saying]. Because to actually make production software is a different skill set than just bootstrapping an app.

Solo.io's Sun adds:

I use [code assistant] Cursor myself, which is one of the best ones. Another is Windsurf. It's enterprise software and we pay monthly licences to share with that. A lot of developers and companies also use it. And the way I'm looking at it, it's just a tool.

Then she adds a crucial point:

A lot of the time, the generated code doesn't actually work! You have to test it – to review the code and make sure it meets your standards. And you have to package it in a way that makes sense.

Honestly, a lot of the time, that multiple iteration, it's still manual today. It might be automated one day, but I don't believe an AI knows what to do. The brain is still inside the human, who must tell the AI what to do, then connect the dots to whatever the environments are. And the testing and validation, and the securing of the supply chain… all that still needs to be done, and to be certified by somebody at the end of the day.

Yes, but what if those skills – hard won by experience – begin to ebb out of the industry? After all, they speak of deep knowledge, of rungs climbed on the career ladder, of real industry insights. What if those rungs no longer exist, or if the current generation of seniors pull up the ladder behind them?

Put simply, what if the vital industry skills needed to check an AI's workings no longer exist at some point in the future? Either because no one bothered to acquire them – or was never given the opportunity to?

Sun responds:

With AI now, you could write a lot faster, but who's going to maintain that code? Can you rely on the AI to maintain that code? So, you have to do more now with the speed provided by the AI.

Good advice, to which Aniszczyk adds:

It will get easier, hopefully, to contribute to open source projects using a lot of these AI tools. But the challenging bit is on the maintainers, who are on the hook for the quality, and let's call it the 'production readiness' of the project, to deal with all this inbound.

We've certainly had some complaints from CNCF staff, and from the leadership, saying 'Hey, we're getting overburdened by random slop coming in.' Or, 'Is there another AI bot or tool that we can install to help us?'

Global collaboration meets geopolitical reality

The open source community remains a beacon of collaboration, of joint endeavour, of international dialogue and partnership. It is a venture that spans continents – a world in which American software engineers, and developers from every continent, can all meet in China – albeit in the relative familiarity of a former British territory – and talk about global software programs and shared visions.

Yet few would deny that the world outside of the CNCF and the Linux Foundation is now far more complex than it was. China is the world's second largest economy, its manufacturing hub, and fully intends to own entire industries, like automotive. Its cities are expanding at a breathtaking rate, and it owns vast amounts of Western debt, though there are cracks beneath that position. And, of course, it is now the intended target of a global trade war, which the US has extended to many of its own allies and friends.

How can such an idealistic venture as the CNCF continue unimpeded? Do the isolationist and nationalist forces that are loose in the world (on both sides) really understand concepts such as open source and global collaboration?

But first, on a simple practical level, how do partners in the US and China even communicate, when some platforms are not open to the other side? Plus, the timezone differences are extreme: for example, Hong Kong is 15 hours ahead of San Francisco.

Sun explains:

LinkedIn is not available for mainland China, which was shocking to me, because I use LinkedIn a lot, and I share my thoughts, my perspectives, and have grown a large follower base on there. So, if you are not plugging into that global, professional networking, it's harder to promote your project. And it's harder to share thought leadership from mainland China to the rest of the world.

Aniszczyk explains that internal conversations often take place on Chinese 'everything app' WeChat, which – on the mainland, but not in Hong Kong – is critical to the day-to-day lives of most connected citizens.

Then Sun adds:

There is a timezone issue that we need to do better for community meetings. I was just talking to [a colleague] and he must be in a meeting at 11pm every Tuesday, which is hard. And honestly, I wouldn't do that. I feel bad because it's ridiculous to ask a member to contribute their time at almost midnight, in their timezone, just to join a community meeting.

Plus, a project started in China tends to have a hard time growing outside of China. Part of this is the language barrier, and part of it the timezone. But there are also the community-building and promotion problems, because those take resources and efforts too. Someone always gets screwed at the end of the day.

Huawei's Wang adds:

I think for contributors, it is still a bit easier than for the maintainers. We also prefer documenting in English and having meeting notes in English. And contributors don't always find it easy to access GitHub or Zoom. But most of the infrastructure is globally available.

So, what of the political perspective: the big picture behind the technical, cultural, and practical challenges? I refer to Trump's America, of course, with its tariffs, and its threats to any US business that offshores jobs and processes.

A wary Aniszczyk says:

Have we noticed any difference in how we work as a community? From a maintainer perspective, no. It's like, 'Do stuff on GitHub, pull requests, review, drink coffee': just the same, same, same as usual. Those tools haven't changed.

But at another level, there's a little bit more complication around regulations and sanctions, things like that. Let's call it 'CNCF back office'. But my role is to ensure that our maintainers never have to think about these kinds of things, regarding regulations and sanctions. We do all the work to ensure that we're compliant.

If anything, it's gotten easier than it was five years ago.

Oh, really? At this point, Sun says:

I think somebody submitted a PR [instruction] around you're not supposed to talk about politics inside of the cloud-native community, as part of our Code of Conduct. I think that really helped, because we're all here to collaborate and to move cloud native forward, even if someone is on a different political side.

Oh, I think we're at the top of the hour.

My take

Yes, I think we are.