Keeping oversight of AI systems aligned with their expanding role inside research pipelines is the core measurement problem for companies running autonomous agents through their own development stack. Anthropic has published three metrics it says AI companies can use to track how fast that development is moving.
The three areas the company said it measured within its own operations are AI-led research and development, oversight of AI agents, and compute allocation. Anthropic framed the metrics as tools for the broader industry.
Each covers a distinct layer. Compute allocation captures where resources are being directed and at what rate, making it a forward signal on which systems are actively expanding. Oversight of AI agents tracks whether human review capacity is keeping pace with autonomous activity. AI-led research and development measures how much of the R&D pipeline the models themselves are now driving.
Anthropic applied all three to its own operations before sharing them as an external framework.