Linear reduces CI wait time to just over 5 minutes, halves runner time
Linear found its continuous‑integration (CI) pipeline lagging as AI‑generated code accelerated development. The issue was raised by CTO Tuomas, who tasked the team with cutting CI costs and latency. The company overhauled its CI system, moving from GitHub Actions to faster third‑party runners, modernizing the toolchain, and rewriting lint rules to avoid TypeScript type checks.
Key points
- PR wait time fell from over 6 minutes to just over 5 minutes after CI overhaul.
- Runner time per test halved and jobs ran 34% faster after switching to third‑party runners.
- Optimizations saved about 87,000 runner‑minutes per month, equal to 11.8% of total CI usage.
The changes delivered measurable gains: jobs ran 34% faster after the runner switch, the native TypeScript compiler (tsgo) cut the weekly median tsc check by 73%, and ESLint lint time fell 68% for API lint and 55% for full‑repo lint. Limiting fetch depth reduced a change‑detection job from 94 seconds to 20 seconds, and a custom checkout action eliminated hangs. Consolidating small jobs and optimizing setup cut per‑shard setup time by roughly 44% (from 110‑140 seconds to 67‑73 seconds). These efficiencies saved about 87,000 runner‑minutes per month (11.8% of total CI usage). Overall, PR wait time dropped from over 6 minutes to just over 5 minutes, and runner time per test was halved, even as the test suite grew fourfold.
Linear notes that the codebase continues to add roughly 2,000 tests each week, so maintaining CI speed will remain an ongoing effort.
AI coding has made CI a bottleneck, so we reworked ours to keep up
linear.app · 21 September 2026
Loading the full article…
This text was published by linear.app. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
1source- Hacker News discussion · 56 pointsnews.ycombinator.com
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us.
More in Enterprise & Industry
All →- Guide proposes AI‑assisted OJT workflow to keep new hires thinking · 1 src
- Schwab bars Anthropic from using advisor data for Claude model training · 2 src
- Lenovo unveils AI-powered SOC model that cuts detection time by 87.5% · 1 src
- Workday launches AI‑powered Workday for Google Sheets add‑on · 1 src
- Salesforce and Siemens launch Agentforce AI solution for industrial sales and service · 1 src
Comments
via GitHub Discussions