DevOps — Culture First, Tools Second

• Culture — shared ownership, blameless post-mortems, devs and ops on the same on-call rotation.
• Automation — CI/CD, IaC, automated testing — reduce toil, eliminate manual handoffs.
• Lean — small batch sizes, continuous flow, work-in-progress limits.
• Measurement — track what matters; see DORA below.
• Sharing — internal docs, blog posts, conference talks; learning travels.

From the multi-year Accelerate research (Forsgren, Humble, Kim). The two top metrics measure throughput; the two bottom measure stability. High performers do well on both — they're not in tension.

• Trunk-based development — small, frequent merges over long-lived branches.
• Continuous Delivery — every green build is deployable; deployment is a business decision, not a technical one.
• Feature flags — decouple deploy from release; ship dark, turn on per cohort.
• Infrastructure as Code — environments are reproducible; no "snowflake" servers.
• Blameless post-mortems — focus on the system, not the individual; assume good intent.
• "You build it, you run it" — devs carry pagers for what they ship.
• Game days / chaos engineering — practice failure on purpose.

• "DevOps team" is a smell. A separate team called "DevOps" recreates the wall it was meant to remove.
• Tools without culture = expensive theater. Buying a CI/CD platform doesn't change behavior.
• On-call burden needs to be deliberate — clear runbooks, tested alerts, sustainable rotation.
• Platform engineering is the modern compromise: a platform team builds an internal developer platform; product teams own delivery on top.
• SRE (Google's variant) emphasizes error budgets and toil reduction; complementary, not opposed.