A developer from OpenAI’s Codex team has criticized the emerging trend of AI agent swarms, describing them as a significant waste of computational resources that fails to improve output quality.
What Happened
According to the developer, deploying multiple agents to work in concert—often referred to as agent swarms—results in excessive token consumption. The critique suggests that the overhead required to coordinate these agents does not translate into better performance or higher-quality results for the end user.
Why It Matters
As the industry explores complex multi-agent architectures to solve difficult coding and reasoning tasks, efficiency and cost remain critical concerns for developers and enterprises. If swarm-based approaches do not yield measurable quality improvements, the increased token usage could make them economically unviable compared to optimized single-agent systems.
The Bottom Line
Internal voices at OpenAI are signaling skepticism toward agent swarms, highlighting that brute-force coordination of multiple agents may not be the most efficient path to improved AI performance.