How I actually use the tools, and where I think most teams get them wrong
One page per tool. Each has the opinion, real configuration, the anti-patterns I have had to unwind, and the Claude workflow that goes with it.
Infrastructure as Code
Terraform
Remote state from day one, modules as components not environments, and a plan every human actually reads.
Read the playbook 7 minPackaging and runtime
Containers & ECS
A container should receive its configuration like a sealed package — no SDK calls at startup, and it runs on a laptop with no cloud credentials.
Read the playbook 7 minDelivery automation
CI/CD Pipelines
The pipeline is the single source of truth for whether a change is safe. Keep the logic in scripts, and keep it under fifteen minutes.
Read the playbook 6 minKnowing what production is doing
Observability
Logs, metrics, and traces answer different questions. Structured logging is also the prerequisite for letting an agent near an incident.
Read the playbook 6 minGlue, tooling, and agents
Python Automation
Most operational work is glue. Write the glue so it is idempotent, observable, and safe to run twice.
Read the playbook 6 minThe reasoning underneath all of these
Every opinion here follows from the same few trade-offs — toil is the enemy, deploying isn't releasing, and a test nobody trusts is worse than no test at all.