OpenAI said it has built Pando, a new Git backend for its monorepo, that handles substantially more pushes and fetches while leaving the way its engineers and AI agents use Git unchanged.
OpenAI · Engineering infrastructure
Pando: OpenAI's new Git backend for its monorepo
OpenAI says Pando handles far more pushes and fetches while leaving the Git interface unchanged for engineers and AI agents. The system is built to scale Git capacity as automated agent activity grows.
~80%
Reduction in push latency
~1 PB
Git data served per day
0
Changes to the Git workflow engineers and agents use
Push latency: before vs. after
Relative scale. An 80% reduction is drawn as 100 units falling to 20. The baseline OpenAI measured against was not disclosed, so the absolute times are unknown.
About 80% less push latency
Why the design matters: same interface, more capacity
Engineers and AI agents
→
Same Git interface
→
Pando backend
→
Capacity grows with agent usage
Keeping the interface fixed means the upgrade happens behind the scenes, without new workflows for users or tools.
Key statement
Monorepo load from AI agents is becoming an infrastructure problem that needs dedicated engineering.
Not yet disclosed:
Pando's internal architecture
The hardware it runs on
The baseline behind the 80% latency figure
Whether it will be released outside OpenAI
Independent verification of the performance numbers
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...