DeepSeek has established an official WeChat account named "DeepSeek Harness Team," operated by a Beijing-based entity verified by Tencent and controlled by DeepSeek [1, 2, 3, 4, 5, 6]. The term "harness" refers to the software scaffolding around large language models that enables AI agents to perform more complex tasks beyond simple responses [1, 2, 3, 4, 5, 6].
The company is actively recruiting for multiple roles on this harness team, which focuses on developing AI agents intended to directly compete with services like Anthropic's Claude Code [1, 2, 3, 4, 5, 6]. On or around August 12, 2026, DeepSeek announced an update to its flagship V4 Pro model, enhancing its agentic capabilities to improve performance and functionality [1, 2, 3, 4, 6].
DeepSeek's move comes amid heated competition between Chinese AI firms and U.S. developers, including Anthropic, OpenAI, Meta, and Nvidia, all vying to create advanced and affordable AI tools for a growing market [1, 2, 3, 4, 6]. Notably, Meta and Nvidia have recently launched open-weight AI models as part of this intensifying U.S.-China AI race [6].
Chinese media reported the formation of the Harness Team account on August 11, 2026, highlighting the company's commitment to expanding its AI agent capabilities [2]. The following day, DeepSeek’s update to the V4 Pro model was publicized through its official WeChat channel, signaling ongoing development progress [1, 3, 4, 6].
DeepSeek’s recruiting efforts continue as it builds out this team, aiming to strengthen its position in the competitive AI agent space.
