GitLab's agent-optimized Git backend cuts AI workload data transfers by 1000x

GitLab reports that traditional Git systems struggle under the load of automated agents, with some tasks requiring up to 10GB of unnecessary data transfers per agent operation. Their internal data shows a 50% increase in GitLab.com code pushes and a 500% growth in codebase sizes over the past year. The company's next-gen SCM solution, unveiled at GitLab Transcend, addresses these issues with server-side queries that reduce agent workloads by up to 1,000x in network traffic during internal tests.