CHAOSS metrics were designed when contributions were produced at human speed, but AI agents and automation are breaking the assumptions behind most of the catalogue. Activity counts (issues, PRs, commits) no longer reliably represent human effort — AI-generated issues inflate numbers while maintainer workload rises. Responsiveness metrics are distorted by bot triage and AI-filed issues. Contributor identity metrics can't distinguish a fleet of agent accounts from real new contributors. Bus Factor and Inactive Contributor metrics miss maintainers who have left but whose tokens still trigger automated merges. Libyears is critiqued as a fundamentally flawed proxy for dependency risk — it ignores unmaintained packages, rewards pinning to abandoned libraries, and can steer teams toward supply-chain-compromised releases. The author argues the CHAOSS Evolution and Risk focus areas need urgent revision, as the core assumption that repository events represent human judgment no longer holds.
