DeepSeek Unveils New AI Framework to Boost Multitasking in Programming Assistants

The initiative focuses on reducing the need for human coordination and supervision when managing several AI agents, as detailed in a research paper by DeepSeek's Harness team, highlighted in Singapore's South China Morning Post on Sunday.

The team emphasized that effective management of multiple agents still requires human input, stating that while single-task limitations hinge on the AI model's capabilities, multitasking constraints largely stem from the user's attention capacity.

According to DeepSeek's publication, an individual can effectively monitor only between three to five active AI agent sessions simultaneously. Their proposed system plans to allocate tasks across different agents and coordinate their activities, allowing humans to concentrate on evaluating outcomes and making decisions.

This development emerges amidst fierce competition in the AI field, with major players like American companies Anthropic and OpenAI, as reported by SCMP.

Last October, Cointelegraph highlighted that Chinese AI models were outperforming their US counterparts in cryptocurrency trading, based on insights from blockchain analytics firm CoinGlass.

DeepSeek's AI chatbot was at the forefront of a crypto trading experiment, achieving the sole positive unrealized return of 9.1%. This outcome caught the attention of cryptocurrency traders, especially considering DeepSeek's development costs were significantly lower compared to those of US competitors.

These advancements signify a shift toward greater AI autonomy in programming, potentially increasing efficiency and reducing the demand for constant human supervision.