Petabytes of Git per Day: Is OpenAI Scaling Engineering or Just Scaling Slop?

Inside OpenAI's Infrastructure Push to Tame an Exponentially Growing Monorepo

OpenAI has published an engineering blog post detailing the technical challenges of scaling Git to accommodate what it describes as petabyte-scale repository data flowing through its systems on a daily basis. The post walks through the architectural bottlenecks that emerge when a monorepo grows faster than off-the-shelf version control was ever designed to handle, covering optimizations around storage, indexing, and distributed access. This comes at a moment when AI labs are under intense pressure to demonstrate operational maturity, not just model benchmarks—and the timing closely mirrors a similar recent infrastructure announcement from Anthropic. Technologically, the piece sits in a lineage of large-scale Git engineering work pioneered by Microsoft (with the Windows monorepo) and Google, where conventional tooling simply breaks under sheer scale. What's notable is that the write-up leans heavily on genuine technical substance rather than the promotional gloss that typically accompanies corporate AI communications.