Running a Green CI Pipeline Isn't the Hard Part Anymore

Description and use

A CI pipeline built for human‑written code becomes a bottleneck when AI generates many commits. Enabling Harness CI intelligence features—dependency caching, selective test runs, build caching, Docker layer caching, test parallelism, and flaky test detection—eliminates redundant work, shortens build times, and improves reliability.

Problem

AI‑generated code increases CI queue length and build times because the pipeline was designed for human‑written code, leading to redundant work such as full dependency downloads, recompilation, and full test suites.

Outcome

The pipeline now skips redundant dependency downloads, recompilation, and tests, cutting build times and improving green‑rate reliability in an AI‑heavy commit environment.

Workflow

• Enable Cache Intelligence by setting cache.enabled:true at stage level to preserve dependency caches.
• Enable Test Intelligence by setting intelligence.disabled:false on test steps to run only affected tests.
• Enable Build Intelligence by setting build-intelligence.enabled:true to reuse unchanged module outputs.
• Configure Docker Layer Caching to avoid rebuilding unchanged base layers.
• Use test splitting and parallelism to run tests concurrently.
• Apply flaky test detection and quarantine to reduce noise from intermittent failures.

Evidence

AI coding assistants removed the slowest part of the development cycle: writing code. What they did not touch is everything after the commit: dependency resolution, compilation, test execution, image builds.

Source

On the first run, the cache step shows "file not found" because nothing has been saved yet. On every subsequent run, it restores the full .m2/repository in seconds.

Source

On the second run, after changing a single file (InputValidator.java), TI selected only 12 tests and skipped the remaining 30. The test run time dropped from 58ms to 16ms.

Source