DeepSeek, a China-based AI company, said on October 31 in a WeChat technical post that its Harness team had built a “super-agent” system to help AI assistants handle multitasking and team workflows with less human coordination.

DeepSeek said the design uses a two-layer architecture, event-driven workflows, asynchronous multichannel interactions, separate storage and computation, and an external watchdog. The system manages session handling and context synchronization in the background.

DeepSeek said an internal instance has been running continuously, helping reproduce and fix software bugs, respond to code reviews, and develop DeepSeek Harness, a plugin-oriented project distributed publicly on GitHub.

According to the post, when team members report software issues in group chats, the system can identify potential defects and try to reproduce and fix them in the background even without being directly tagged.

The post adds to an industry push toward more autonomous software agents. OpenAI has described an agent-first approach to engineering workflows.

If offered to the public in China, such generative AI services could fall under the country’s generative AI rules, which include safety, data-source, and content-control requirements.

🔗 Source: Yicai

Recent DeepSeek developments