🚀 Excited to announce the launch of Trunk Flaky Tests! You can now detect, quarantine, and eliminate flaky tests from your codebase. Works with any language, any test runner, and any CI provider. Keep your CI Green. Get started today at https://t.co/pZDn0y2At6

@ProductHunt Ready to give it a try? 📺 Watch the demo: https://t.co/B0a77v7f07 🚀 Vote on Product Hunt: https://t.co/2HMdwx5aZc 🌐 Learn more: https://t.co/MGMgxc0mKP 📖 Get started:

✅ Works with any language and CI provider ✅ Quarantines flaky tests on CI ✅ Keeps your team shipping instead of debugging random failures We're live on @ProductHunt today - every upvote helps us reach more frustrated developers! 🔗

The problem: You push code, CI fails on some unrelated test, you rerun it and it passes. Sound familiar? The solution: Automatically detect and quarantine flaky tests so they can't block your team's progress.

🎉 Trunk Flaky Tests is now out of beta after months of user testing and feedback! No more random CI failures blocking your PRs. No more "just rerun it and pray it works" moments. Watch how it works https://t.co/B0a77v7f07

The math is simple: - Spending $3K+ monthly on CI? - Have 20+ engineers? - 30+ PRs per day? Trunk pays for itself in CI savings alone. 📊 Read More: https://t.co/ImcF3KmJ3K 🚀 Learn more about Trunk Merge Queue: https://t.co/mV95TR2ivt 5/5

Real customer example: - Team of 50 engineers - Was spending: $15K/month on CI - After Trunk: $4K/month on CI - Savings: $11K/month vs $500 Trunk cost ROI: 22x in CI costs alone (not counting dev productivity gains) 4/5

Trunk's batching approach: - Batch [PR1 + PR2 + PR3]: 10 min test ⏱️ Total: 10 minutes, 1 CI run Cost savings: 70% reduction If batch fails? Smart bisection finds the culprit automatically. 3/5

Traditional merge queue approach: - PR 1: 10 min test ⏱️ - PR 2: 10 min test ⏱️ - PR 3: 10 min test ⏱️ Total: 30 minutes, 3 CI runs Cost: $$$ for each individual test This doesn't scale. 2/5

💸 Your CI bill is growing faster than your team, and it's not sustainable. Here's why most merge queues are expensive... and how to fix it: 🧵👇

The result: ✅ Teams don't block each other ✅ Fast changes merge in minutes, not hours ✅ Complex changes get proper validation ✅ Monorepo becomes productivity multiplier again Stop fighting your architecture. Make it work for you. 📖 https://t.co/ImcF3KmJ3K 4/4

Trunk's solution: Parallel queues 📝 Docs queue: 2 min → merge ✅ 🔧 Backend queue: 45 min → merge ✅ 🎨 Frontend queue: 15 min → merge ✅ All run simultaneously. Teams work independently again. 3/4

Classic monorepo problem: 📝 Docs team: 2-min PR stuck behind... 🔧 Backend team: 45-min database migration 🎨 Frontend team: 15-min feature addition Everyone waits for the slowest PR. Independent teams blocked by unrelated changes. 2/4

Your monorepo is supposed to make development faster. Instead, it's become a coordination nightmare. Here's how to fix it: 🧵👇

The numbers don't lie: - 3-5x faster merge times - 60-80% lower CI costs - Near-zero main branch breakages If you're spending too much on CI or have 20+ engineers, the ROI is obvious. 📖 Full comparison: https://t.co/ImcF3KmJ3K 6/6

Enter Trunk Merge Queue: 🚀 Parallel queues (independent changes don't wait) 💰 Batching (90% CI cost reduction) ⚡ Optimistic merging (fast PRs don't wait for slow ones) 🔍 Full visibility and debugging 5/6

GitHub Merge Queue limitations: - Single queue = bottlenecks - No batching = high CI costs - Limited visibility when things fail - Queue jumping rebuilds EVERYTHING For small teams? Fine. For 50+ engineers? You need something better. 4/6

GitHub's merge queue fixes this with "predictive testing": ❌ Old way: Test PR against stale main ✅ New way: Test PR against "future main" (including other PRs ahead in queue) But GitHub's implementation has limits... 3/6

9:00 AM - Alice creates PR: renames foo() → bar() 9:05 AM - Bob creates PR: adds code calling foo() Both PRs test against OLD main branch. Both PRs pass ✅ Both PRs merge. Main branch: 💥 BROKEN This is the "stale CI problem" 2/6

🧵 Your CI is lying to you, and it's breaking production. Here's the timeline that destroys main branches everywhere: 👇 1/6