DeepSeek Develops AI Agent System Requiring Less Human Input

DeepSeek unveiled an AI agent design that coordinates multiple coding tasks with less human oversight, following its earlier lead in a crypto trading experiment.

maisiekooc
Maisie Morrison

AgentLocker Editor

AI News
DeepSeek Develops AI Agent System Requiring Less Human Input

Chinese artificial intelligence company DeepSeek has revealed a new design for AI coding assistants. The system aims to let AI agents handle multiple tasks with less help from people.

The plan was laid out in a paper published by the company’s DeepSeek Harness team. The South China Morning Post reported on the paper on Sunday, November 1.

The goal is to cut the amount of human work needed to coordinate and watch over several AI agents at once.

The design targets AI coding assistants. These are tools that help write and manage software code.

How the DeepSeek System Would Work

According to the paper, managing many AI agents takes a lot of human attention. The team wrote that “single-task limits depend on model capabilities, while multitask limits depend on human attention.”

In other words, how well an AI handles one task depends on the model itself. How many tasks can run at once depends on how much a person can keep track of.

The paper said one person can only oversee three to five active AI agent sessions at the same time. Beyond that, it becomes hard for a human to keep up.

DeepSeek’s proposed system would spread tasks across different agents. It would also coordinate the work those agents do.

This setup would leave people free to focus on reviewing results and making final decisions. The ongoing management of the agents would be handled by the system.

DeepSeek’s Record in Crypto Trading

DeepSeek has drawn attention in the crypto market before. In October 2025, Cointelegraph reported that Chinese AI models were outperforming US models in cryptocurrency trading.

That report was based on data from blockchain analytics platform CoinGlass. It looked at how different AI chatbots performed when trading crypto.

DeepSeek led the trading experiment. It was the only AI model in the test to generate a positive unrealized return, which came in at 9.1%.

The results surprised many crypto traders. DeepSeek was built at a fraction of the cost of its US rivals.

Competition among AI developers remains intense, according to the SCMP. DeepSeek is competing with US companies including Anthropic and OpenAI.

The new paper is DeepSeek’s latest move in that race. It focuses on coding assistants that can work with less human oversight, as reported by the SCMP on Sunday.

From our research desk
AI Jobs Automation Index
Which jobs are AI tools targeting most? We mapped 3,400+ AI tools to real job functions — with BLS employment & salary data.
Explore the index
maisiekooc

Written by

Maisie is a news writer at Agent Locker, covering the latest developments in artificial intelligence, emerging technology and the companies shaping the future.

Discover AI Agents