DeepSeek is presenting a prototype of a more autonomous programming agent in an effort to compete with Anthropic’s Claude Code. With this move, the Chinese AI model developer is making a more pronounced entry into the market for programming assistants.
DeepSeek calls the new open-source project Super-Agent. The prototype can independently monitor group chats among developers to identify programming issues and suggest solutions, even when developers do not ask for them. At least, that is what the DeepSeek Harness team reports in a post on WeChat. The new agent is designed to require as little human attention as possible while providing as much value as possible, DeepSeek explains in a GitHub post.
DeepSeek founded the Harness team earlier this year. The team develops AI agents capable of competing with programming services from leading U.S. AI companies, including Anthropic.
The term “harness” refers to the software infrastructure built around a large language model that enables AI agents to perform more complex tasks. Anthropic and other developers have designed “harnesses” to help their software perform better at streamlining programming work and other tasks.
Fast-growing market for AI agents
Since early last year, when DeepSeek made its international breakthrough with the R1 model, Chinese companies have been locked in fierce competition with U.S. developers to build more advanced and less expensive AI tools. Although OpenAI’s ChatGPT kicked off the global AI race, many companies have since shifted their focus to the lucrative and rapidly growing market for AI agents that can help automate work for developers and business professionals.
To bolster its war chest and compete with Anthropic and OpenAI, DeepSeek raised approximately 100 billion yuan ($15 billion) in its latest funding round, ahead of its initial public offering. It was one of the largest funding rounds ever for a Chinese company.