DeepSeek unveils new AI that autonomously performs multiple tasks
DeepSeek unveils a new AI design that allows multiple digital assistants to collaborate independently. The Chinese company is thus further increasing competition with OpenAI.
Chinese AI company DeepSeek has unveiled plans for a new generation of AI assistants who can work much more independently. The system must be able to execute several commands at the same time without users having to intervene constantly. With this, DeepSeek is further increasing its competition with American companies such as OpenAI and Anthropic.
The research team behind DeepSeek Harness presented a new design for a so-called Super-agent last weekend. Currently, AI assistants can already program independently and carry out complex assignments. Nevertheless, users often have to control various AI systems themselves and monitor their progress.
According to DeepSeek, one person can only oversee an average of three to five active AI sessions at a time. As a result, not only the technology, but above all the available human attention is a limitation.
The new system needs to change that. A central AI assistant distributes jobs across multiple specialized AI systems, which then work independently on different tasks. As a result, users no longer have to guide every step. They can focus on monitoring results and making important decisions.
For the time being, this is a research design with an experimental prototype, not a fully rolled-out product.
The plans come at a time when competition between Chinese and American AI companies is becoming increasingly fierce.
Companies such as OpenAI and Anthropic are also developing AI assistants that can write software independently and perform complex work.
DeepSeek has previously attracted attention by developing powerful AI models at a relatively low cost. The Chinese company also made itself heard within the crypto sector. During an experiment with automatic trading in crypto coins, which was reported in October 2025, DeepSeek was the only participating AI model to achieve an unrealized positive return of 9.1 percent.
It was a limited test and not proof that DeepSeek can invest structurally better than other AI systems. With its new design, DeepSeek wants to take the next step: AI systems that not only carry out individual assignments, but also collaborate independently on larger projects.