Gemini 4 Argon refactors 6M-line Chromium folder
Self-reported figures say Gemini 4 Argon refactored a 6M-line folder in Chromium's 40M-line codebase in weeks instead of months, found 16+ timeout and race-condition bugs across 700+ internal labs, and saved 955 engineering hours.
gemini 4 argon helped refactor a 6-million-line folder inside chromium's 40-million-line codebase. work estimated in months took weeks. that is the frontier-model test i care about: not one clever function, but changing old code without losing the behavior hidden across millions of lines. the same report says argon found 16+ timeout and race-condition bugs while migrating 700+ internal labs, saving 955 engineering hours. it also helped raise gemma 4 inference from 37.9k to more than 108k tokens/sec. all of those numbers are self-reported, but they point at where coding agents are going: repo-scale maintenance, not autocomplete.