Chinese AI firm DeepSeek announced a new framework intended to make AI coding assistants more autonomous, allowing them to handle several tasks with far less human input. The proposal, detailed in a paper by the company’s DeepSeek Harness team, suggests a system that can allocate work across multiple agents and coordinate their efforts, leaving users to simply review outcomes and make final decisions.

Reducing Human Attention Bottlenecks

According to the DeepSeek paper, a person can effectively supervise only three to five active AI agent sessions at once. The authors argue that while single‑task limits are bound by model capabilities, multitask limits are constrained by human attention. By automating task distribution and coordination, the proposed architecture seeks to free human operators from the constant monitoring required when juggling several agents.

Competitive Landscape

The announcement arrives amid heightened rivalry among AI developers, notably US‑based Anthropic and OpenAI, as reported by the South China Morning Post. DeepSeek’s push for greater autonomy reflects a broader push to close the gap with Western rivals, especially after earlier reports highlighted Chinese models outperforming US counterparts in cryptocurrency trading.

Previous Crypto‑Trading Performance

Cointelegraph noted that in a prior experiment, DeepSeek’s chatbot was the sole AI model to generate a positive unrealized return of 9.1% on crypto assets, according to data from analytics platform CoinGlass. The result surprised traders given DeepSeek’s development costs were reportedly a fraction of those of its US competitors.

What to Watch Next

Observers will monitor whether DeepSeek can translate its multitasking framework into a commercial product and how quickly it can be adopted by developers. Attention will also focus on regulatory responses, especially in the United States, where Chinese AI technologies have faced increasing scrutiny.