Chinese Syndrome: Vibe coding may be reshaping how software gets built, fueling billion dollar valuations in Silicon Valley. But behind the hype, many of these groundbreaking tools may rely on core AI models developed in China – a detail their creators are not openly discussing. That tension between innovation, national tech pride, and transparency is now at the center of one of AI's fastest growing trends.
Vibe coding is becoming one of the most popular use cases for generative services and AI models, driving up the valuations of several US companies. However, some of these models originate from China, which has raised concerns among some observers.
Cognition and Cursor, two US companies at the forefront of the vibe coding movement, have introduced their own tools that turn software programming into an agentic process. Both claim to have achieved unprecedented performance and efficiency, but they have not disclosed their apparent reliance on advanced AI models developed in China.
Cognition recently launched SWE 1.5, its latest AI model designed and optimized for software engineering. The company says it was trained on hundreds of billions of parameters and can nearly match state of the art performance. SWE 1.5 is meant to address the tradeoff between speed and code quality that often affects AI text generation services.
According to Cognition, the agent technology uses "a leading base model" released under an open source license, although no specific details have been provided. When users asked the tool about its foundation model, it appeared to reference the GLM series developed by Zhipu AI, a company based in Beijing.
Zhipu AI confirmed the speculation, suggesting that SWE 1.5 may be based on its GLM 4.6 model. Cognition has not commented on the matter. The company is based in the San Francisco area, is valued at 10.2 billion dollars, and aims to deliver the fastest coding agent experience available.
Anysphere did not reveal what model Composer uses, but users noticed that it occasionally produced text in Chinese.
Anysphere is another developer of AI coding services with an exceptionally high valuation. The company recently introduced Composer, an agent designed to support vibe coding projects in Cursor. Cursor is the best way to write code with AI, the company says, and Composer can generate programming output four times faster than similar models. Anysphere did not reveal what model Composer uses, but users noticed that it occasionally produced text in Chinese. Like Cognition, Anysphere is based in San Francisco and has tripled its valuation to nearly $10 billion dollars in just six months.
Neither Cognition nor Anysphere or Cursor have commented on the Chinese origin of the models powering their products. Zhipu AI said that US companies adopting its GLM models is not necessarily negative, as it could benefit open source contributions across the AI ecosystem.
According to Florian Brand, a PhD student at Trier University in Germany, the origin of the base model used by US companies is not the most critical issue. Creating effective AI agents requires an additional phase of fine tuning, which he says is the real secret behind the success of these ventures.