AI Agents Assemble Internet Measurement Tools Independently to Resolve Disruptions

The new ArachNet system demonstrates that AI agents based on large language models can independently assemble internet measurement processes in the same way as human experts.

According to researchers, measuring the functioning of the internet has reached an accessibility crisis. When investigating network disruptions, complex analyses are needed quickly: infrastructure must be mapped, routing examined, and dependencies between different services modeled. This currently requires the combination of several specialized measurement tools and deep network expertise.

ArachNet approaches the problem by assembling an expert-like workflow from several separate AI agents. The system includes four specialized agents that reflect the work of a network expert, from breaking down a problem into individual questions to planning and executing the necessary measurements.

The researchers' key claim is that expertise in network measurements follows predictable, repetitive combination patterns. According to the system, these patterns can be modeled so well that AI can select and combine existing measurement tools without continuous human guidance.

According to the research, ArachNet is the first system to demonstrate that such an agent-based approach works in internet measurements. The goal is not to replace network experts but to reduce routine manual work and make complex measurement processes more accessible, for example, to network administrators during disruptions.

If the approach becomes widespread, it could speed up the diagnosis of network faults and lower the skill threshold for demanding measurements—especially in situations where every minute of downtime costs.

Source: Towards an Agentic Workflow for Internet Measurement Research, ArXiv (AI).