
High Bit Trunk: Fixing CI at Scale (Merge Queues, Flaky Tests, and Shipping Code)
Mar 10, 2026
Eli Schleifer, founder and CEO of Trunk who built developer infrastructure at Microsoft, Google, and Uber, talks CI reliability at scale. He covers why CI becomes a bottleneck as teams grow. He explains merge queues, flaky tests, batching and anti-flake tactics. He also discusses AI-driven fixes, dynamic parallelism, and the build vs buy tradeoffs for developer tooling.
AI Snips
Chapters
Transcript
Episode notes
Build Resilience Around External Flakiness
- Real-world orchestration must be resilient to unreliable external systems like GitHub webhooks.
- Trunk builds code to detect and surface when failures are due to upstream services so teams know the true cause.
AI Increases PR Volume And Merge Queue Need
- AI and agentic coding tools increase PR volume and conflict risk, driving higher demand for merge queues.
- As bots generate more changes, merge queues protect against stale CI data and landing-time conflicts.
Flaky Tests Quietly Devour Productivity
- Flaky tests are a "vampiric" drain: false negatives make engineers waste time debugging non-issues.
- Trunk records test history, classifies flakiness, quarantines flaky failures, and lets engineers proceed when only flaky tests fail.

