DeepSeek advances AI agent autonomy in competitive coding assistant market

Traders Union
All News
Crypto
DeepSeek AI agent advances

Highlights

DeepSeek's Harness team proposed a system allowing AI coding assistants to autonomously manage multiple tasks, reducing required human oversight to three to five sessions per person.
The new setup aims to centralize agent coordination, responding to increased competition from Anthropic and OpenAI and reflecting industry efforts to optimize productivity and scalability.
In cryptocurrency trading experiments cited by Cointelegraph, chatbot DeepSeek achieved a 9.1% unrealized return, outperforming more expensive U.S. AI models.

Proposed system targets lower human oversight

As reported by Cointelegraph, citing the South China Morning Post, DeepSeek's Harness team says it is developing a design to increase the autonomy of AI coding assistants so they can handle multiple tasks with less human intervention.

In a paper published by the Chinese artificial intelligence company, the team says managing several agents at once still requires significant human attention. It argues that "single-task limits depend on model capabilities, while multitask limits depend on human attention," highlighting a bottleneck in how developers coordinate and monitor agent-based systems.

The paper says one person can oversee only three to five active AI agent sessions simultaneously. Under the proposed setup, tasks are distributed among agents and their work is coordinated centrally, while humans remain responsible for reviewing results and making final decisions.

Pressure builds as AI competition intensifies

The proposal emerges as competition remains intense among AI developers, including U.S. companies Anthropic and OpenAI. The focus on reducing manual oversight reflects a broader industry push to improve productivity and scale the use of AI agents in software development workflows.

DeepSeek has also drawn attention in crypto-related AI applications. Cointelegraph reported in October last year that Chinese artificial intelligence models outperformed U.S. counterparts in cryptocurrency trading, citing data from blockchain analytics platform CoinGlass.

In that experiment, chatbot DeepSeek was the only AI model to post a positive unrealized return, at 9.1%. The result stood out because the model was developed at a fraction of the cost of its U.S. rivals, surprising some crypto traders.

In our earlier report on CEO adoption of AI agents, we highlighted that many large U.S. companies are seeing efficiency and growth gains, but few leaders consider the technology truly transformational yet. The survey also pointed to continued investment in AI infrastructure and a focus on hiring and retraining, underscoring that integrating agentic systems requires operational redesign rather than a simple plug-and-play rollout.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.