Monorepo surgery, CI discipline, semantic layers, and cost levers — Discord's petabyte dbt to Datadog's 5,000-person self-serve.
Past a few thousand models, dbt stops being a tool and becomes a platform you engineer: Discord's petabyte-scale surgery on dbt internals, Checkout.com's 27-project CI library, the Airbnb/DoorDash/Netflix semantic-layer canon, and the SQLMesh-era platform decisions.
The trust pipeline: CI as a platform, Slim CI's deferral asymmetry, the data-vs-unit-test taxonomy, and contracts that fail at compile time with ownership-routed alerts.
The finance model broke on a Tuesday. The PR that broke it had passed CI — green checks, approvals, the works. Post-incident review finds the chain: CI "built the changed models" against production-sized tables and timed out, so someone had marked the job non-blocking months ago; the model's only test asserted non-null on a column the change didn't touch; and the upstream engineering team had renamed a source column Friday — which nothing they own would ever catch, because nothing they own knows your warehouse exists. Three gaps, one incident. Which gate was supposed to be where?
The full week 2 brief is part of LeetData Pro.
