Every developer has experienced the pain of a broken main branch. You merge your PR, everything looks green, and then—CI fails. Someone else merged a conflicting change seconds before you.
The Problem with Traditional Merge Queues
Traditional merge queues solve this by serializing merges. But this creates a new problem: throughput. When you have dozens of PRs ready to merge, waiting in a queue becomes its own bottleneck.
Speculative Execution
Our new approach borrows from CPU design: speculative execution. We don't wait for each PR to merge before testing the next one. Instead, we predict the likely merge order and test PRs against speculative future states of main.
- PRs are tested against their predicted post-merge state
- Conflicts are detected before they reach the queue
- Failed predictions are rolled back and retried automatically
- Throughput increases by up to 5x compared to serial queues
It's like having a time machine for your CI pipeline.
Getting Started
Speculative merge queues are available today for all Pro and Enterprise customers. Head to your repository settings to enable this feature, or reach out to our team for a guided setup.