Chinese AI Researchers Are Moving to X, and It Is Changing How the West Reads China's AI Boom

Key points

• Around 30 Moonshot AI staff, including two cofounders, had active X accounts as of mid-2025.
• Chinese researcher Meng Fanqing says he began posting actively on X in mid-2025, after DeepSeek's R1 model went viral worldwide at the start of the year.
• Several DeepSeek employees whose passports were reportedly confiscated by the Chinese government still post regularly on X.
• Independent AI analyst Tiezhen Wang says OpenAI and Anthropic researchers have grown noticeably quieter on social media as their companies have scaled.
• Domestic Chinese platforms Zhihu and Xiaohongshu are seen by technical researchers as too superficial or too light on serious discussion for AI topics.

For years, Western observers assumed China's AI labs were a black box. They were wrong, and the proof is a quick scroll through X.

Over the past year, researchers from some of China's most prominent artificial intelligence companies have joined X in growing numbers, posting about their work, debating ideas with Western colleagues, and building public profiles. Wired AI first reported the trend in detail, drawing on interviews with researchers across the Chinese AI community.

Who is actually posting?

Moonshot AI, the Beijing-based startup behind the Kimi K3 reasoning model, leads the pack. By late June 2025, roughly 30 people claiming current Moonshot affiliation had active X accounts, including two of the company's cofounders. Staff from Minimax and Z.ai are also visible, and even some employees at DeepSeek, which made headlines globally when its R1 model (a reasoning system that matched top Western models at a fraction of the cost) went viral in early 2025, post regularly despite reportedly having their passports held by Chinese authorities.

Meng Fanqing, cofounder of startup Evolvent AI and a former Moonshot intern who worked on the Kimi K2.5 model, puts it plainly. "X is the best community" for AI discussion right now, he says, citing its recommendation algorithm and its mix of Chinese and international audiences.

Why not stay on Chinese platforms?

Chinese platforms simply do not offer the same environment for serious technical exchange.

Zhihu, a question-and-answer site once compared to Quora and popular with professionals, has shifted heavily toward fiction and entertainment content since 2020. Expert contributors have drifted away. Xiaohongshu, known in the West as RedNote and popular with younger urban users, attracts a less technically focused crowd. When Moonshot published a notable research paper in March 2025, one current researcher (who spoke anonymously because they were not authorised to talk to media) said Chinese platform readers fixated on the authors' ages and family backgrounds, while foreign readers actually discussed the science.

What does this mean for Western readers of AI news?

It narrows the gap between what Chinese labs are doing and what the rest of the world understands about them.

Tiezhen Wang, an independent AI analyst and former researcher at Hugging Face (the company behind one of the most widely used open-source AI tool libraries), says researchers at OpenAI and Anthropic have grown quieter as those companies have become large and protective of their methods. Chinese researchers are stepping into that attention vacuum.

Wang, who advised Moonshot in a one-off consultation in 2025, says there is a commercial logic too. "The AI industry works just like the finance industry now," Meng says. "Each individual and company needs their own branding." A researcher with a following on X gives their employer reach that a corporate press release never could.

For ordinary readers, the practical upshot is straightforward: following researchers directly, rather than waiting for translated summaries, is now a realistic way to track what Chinese AI labs are actually building.

"People have a lot of wrong hallucinations about what China is doing," Wang says. "X is the door for them to better understand China."

Common questions

Does this mean Chinese AI research is now fully open?

No. Companies like DeepSeek still keep core training methods and model architectures private. What has changed is that individual researchers now share context, paper discussions, and professional opinions publicly, which is more than was visible before.

Should I trust what these researchers post?

Treat their posts as you would any source: useful colour and context, not official company positions. As with Western researchers, individual views do not always reflect employer strategy, and researchers do have an interest in building their own public profiles.