Update from Hell tests whether coding agents can survive dependency upgrades that break code silently
arXiv·medium signal
arXiv 2608.30300 (2026-08-31) targets the case where a dependency upgrade changes function signatures, types, APIs or runtime semantics without those changes being communicated to downstream maintainers, and asks whether coding agents can perform the required source adaptations. Dependency bumps are one of the most commonly delegated agent tasks and one of the least verified, since the tests that would catch the breakage are often the thing that broke.