Technology
China's AI app market: A few leaders, a fast-moving race
Read more
The one I find most interesting at the moment is China’s growing focus on AI safety. Over the past several months, we have seen a wave of new safety guidelines and regulations coming out — and what is notable is that this is increasingly a government-driven priority, not just something left to the industry. China is signaling that it wants to be a serious actor on AI governance, not only on AI development. That is a meaningful shift, and I think it will generate a lot of developments worth watching in the near future.
What frustrates me most is how extreme the narratives are — on both sides. In China, in the U.S., you constantly hear: we are winning, they are falling behind, we are ahead, they are catching up. And most of these narratives are simply not objective. They are not grounded in evidence. The actual picture is much more nuanced and much more interesting than either narrative suggests. That gap between the discourse and the reality is actually one of the things that makes this work meaningful.
Geopolitics, I think. That is the thread. Whether I’m looking at AI chips, medtech, or machinery, geopolitics is always in the background, and often in the foreground, shaping who can sell where, which suppliers are trusted, and what rules apply. What I find genuinely exciting is following how those dynamics play out in real markets. AI is probably the most visible geopolitical battlefield right now, but it is not the only one.
You need a broad set of sources — news, social media, podcasts, Chinese online communities, and you need to check them constantly. But the bigger challenge is not finding information. It is filtering it. In AI, the volume of new releases, claims, and announcements is enormous, and a lot of it is noise. The ability to distinguish what is actually significant from what is hype is, in my view, the core skill. That requires knowing the field well enough to evaluate claims critically, and knowing which sources tend to be reliable and which tend not to be. That knowledge takes time to build.
I keep coming back to one thing: the ability to separate signal from noise. This is a field where there is genuinely too much information, where things change every week, and where a lot of what gets published — especially around new model releases or capability claims — is overstated or incomplete. Good analysis means not being swept along by that. It means being rigorous about what the evidence actually says, and being honest when the picture is still unclear.
