The term Vibe Coding (agentic programming) is reshaping how software is built, driving multi-billion dollar valuations in Silicon Valley. However, behind the hype, many of these breakthrough tools may be relying on core AI models developed in China—a detail their founders have not publicly discussed. This tension between innovation, national technological pride, and transparency is currently at the center of one of the fastest-growing trends in the AI sector.
Vibe Coding is becoming one of the most common use cases for generative services and AI models, driving the valuations of several US companies to sky-high levels. However, the Chinese origins of certain foundation models are raising concerns among observers.
Cognition and SWE 1.5 suspected of being based on Zhipu AI’s GLM model
Cognition and Cursor, two US companies leading the vibe coding movement, have introduced their own tools aimed at turning software programming into an agentic process. Both claim to have achieved unprecedented performance and efficiency, yet they have not disclosed their clear dependence 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 stated the model was trained on hundreds of billions of parameters and is nearly capable of matching state-of-the-art performance. SWE 1.5 is expected to resolve the trade-off between speed and code quality that often affects AI text generation services.
According to Cognition, their agentic technology utilizes “a leading base model” released under an open-source license, though no specific details were provided. However, when users asked the tool about its foundation model, it appeared to reference the GLM series developed by Zhipu AI, a Beijing-based company.
Zhipu AI has confirmed the speculation, suggesting that SWE 1.5 may be based on their GLM 4.6 model. Cognition has not commented on the matter. The San Francisco-based company is valued at $10.2 billion and aims to provide the fastest coding agent experience currently on the market.
Anysphere’s Vibe Coding and evidence of leaked Chinese language
Anysphere is another AI coding service developer with an extremely high valuation. The company recently introduced Composer, an agent designed to support vibe coding projects in Cursor. Cursor is described by the company as the best way to write code using AI, and Composer can generate programming output four times faster than similar models. Anysphere has not disclosed the model that Composer uses, but users have noticed that it occasionally generates text in Chinese. Like Cognition, Anysphere is based in San Francisco and has tripled its valuation to nearly $10 billion in just six months.
Neither Cognition, Anysphere, nor Cursor have commented on the Chinese origins of the models powering their products. Zhipu AI stated that US companies adopting their GLM models is not necessarily negative, as it can benefit open-source contributions across the AI ecosystem.
According to Florian Brand, a PhD student at the University of Trier in Germany, the origin of the foundation models used by US companies is not the most important issue. Creating effective AI agents requires an additional fine-tuning stage, which he believes is the real secret behind the success of these projects. However, this lack of transparency raises significant questions regarding the independence and control of core AI technology in the US.
