A software engineer at WHOOP adds a new column to a Postgres table on Monday morning. By that afternoon, the column is queryable in Snowflake and available to any downstream model or dashboard, alongside every other column on that table. Nobody filed a ticket. No data engineer touched anything.
Change Data Capture (CDC) is the pattern that makes this possible: every insert, update, and delete on a source table is streamed out as an event, and downstream systems consume that stream to stay in sync. That workflow is the point of the CDC platform we run today, and getting there meant deliberately avoiding the streaming-upsert pattern most CDC-to-lake pipelines converge on.