Airflow and Dagster both schedule and run data pipelines, from different starting points.
Airflow starts from tasks: a Dag is a graph of steps with a schedule. It has run in production since 2015, has providers for most databases, clouds and SaaS tools, and is offered as a managed service by AWS, Google Cloud and Astronomer. Airflow 3 added assets and @asset, so data-aware scheduling is first class, but the unit you operate is still the Dag run and its tasks.
Dagster starts from assets: you declare the tables and files you want to exist and how each is computed, and the graph follows from their dependencies. Partitions, asset checks and the lineage view are central, and local testing is built in.
Neither is better in general. Airflow fits teams with many systems to coordinate, existing Airflow knowledge and a need for broad provider coverage. Dagster fits teams who think in data products, want lineage and data quality in the same tool, and are starting fresh. The Dagster course in The Work Gym covers the other side of this comparison.
The same table, described from each side.