DOP's core runs on Cloud Run and on a k3s cluster with no change to the domain. What makes that true is not the folder layout: two adapters per port from day one, one contract suite every adapter passes, and a test that breaks the build when the domain imports infrastructure — with the code.
Read the article →A delivery architecture that survived several apps, two rings and more than one cluster: a versioned blueprint library, thin app pipelines, a fleet repository that is the only thing talking to Kubernetes, and a live ring that releases what preview already ran instead of building again.
An agent opens a pull request in twenty minutes; a person takes an hour to review it. DOP's answer: no green, no PR — and green has to be earned on a commit, not a dirty tree.
Twenty-three years building and running systems where failure is expensive — across stacks, teams and business domains that had little in common except the need to work.
Bringing AI into how a team builds software — agent-assisted engineering with the specifications, review gates and delivery cadence that make it dependable.
AWS and GCP environments designed for the load they will actually carry, automated end to end, with the trade-offs written down rather than assumed.
From the first specification to a shipped product: architecture, the team, and the operating cadence that makes delivery predictable instead of heroic.
There is more to life than work.
