1.2 million stars! Developers lined up to "submit their resumes", and DeepSeek's internal beta post turned into a thorough investigation of Agent's open source

AI Application Wind Vane (Public Account: ZhidxcomAI) Author | Bi Weihao Editor | Mo Ying Zhidixi reported on August 4 that on August 1, Cui Tianyi, the leader of the Agent Harness team of DeepSeek, issued a document to recruit DeepSeek Harness internal testers, requiring applicants to have participated in the establishment and maintenance of open source Agent projects, and to submit the GitHub warehouse as a representative work. An internal beta recruitment post quickly turned into a survey of the Agent open source ecosystem. A large number of developers poured into the comment area with their own projects, ranging from personal harnesses and coding agents to memory systems, security tools, and evaluation frameworks, covering almost all directions of current agent infrastructure exploration. According to developer statistics, as of 11:30 a.m. on August 3, there were 769 applicants, 712 open source warehouses after deduplication, a total of more than 1.2 million Stars, spanning 18 tracks, and the comment area instantly became an "Open Source Agent Roadshow". Statistics warehouse address: https://github.com/Octo-o-o-o/DeepSeek-harness-applicants Judging from the current public information, DeepSeek has used Harness internally to complete the benchmark test of the official version of DeepSeek-V4-Flash, and the community has also begun to speculate that DeepSeek may build an AI Coding for software engineering scenarios in the future. Agent. According to rumors circulating in the community, this product may have the capabilities of independent planning, tool calling, code execution, etc., and will add mechanisms such as Memory and project warehouse awareness (Repo Awareness) around the long-cycle Agent workflow. The positioning may be benchmarked against existing Coding Agent products such as Claude Code and Codex. This information has not yet been officially confirmed by DeepSeek. However, judging from the Harness benchmark results previously announced, De