CI/CD pipeline
The automated chain that builds and tests every push, then delivers or deploys whatever qualifies, so nobody memorizes the steps.
See it
What it is
A pipeline is an ordered chain of automated jobs that a code event triggers: push a commit, and a machine somewhere checks out the repo, installs dependencies, runs lint and tests, builds an artifact, and (if everything is green) deploys it. CI is the build-and-test half. CD is the ship-it half, and it means one of two things depending on the team: continuous delivery, where every passing change is kept ready to release and a human presses the button, or continuous deployment, where every passing change goes straight out. Plenty of pipelines validate every push but only deploy from certain branches, certain tags, or after an approval, and that is still CD. Most people say 'CI/CD' because it all lives in one YAML file: GitHub Actions, GitLab CI, CircleCI, Jenkins.
Reach for one the moment more than one person touches the repo, or the moment you catch yourself deploying by hand and hoping you remembered the build step. The pipeline is the written-down version of what your best developer does from memory on a good day.
Gotcha: pipelines rot into the slowest part of your day. A 20 minute run on every commit trains people to stop waiting for it. Cache dependencies, run jobs in parallel, and only rebuild what the change touched. Second gotcha: the runner has none of your local machine's secrets or Node version, so 'works on my machine' fails there first.
Ask AI for it
Write a GitHub Actions CI/CD pipeline for this repo. Create separate parallel jobs for lint, typecheck, tests, and build (steps inside a single job run in sequence, so parallelism has to come from jobs), each doing its own checkout, runtime setup, frozen-lockfile dependency install, and dependency caching. Add a deploy job that depends on all four and runs only on merge to main, reading credentials from repository secrets (never inline them). Pin action versions, set a sensible job timeout, and add concurrency so a new push cancels the stale run. A workflow cannot make its own checks required, so finish by giving me a GitHub ruleset or the gh api calls that require those exact check names on main.