A Job is a DAG of tasks, not one giant notebook
On the canvas
Four tasks on the canvas. Each carries its own type chip: NB for notebook, DBT for dbt, SQL for a query task.
In plain English
A Databricks job holds a list of tasks, and each task names its own type — notebook_task, spark_python_task, python_wheel_task, pipeline_task for a DLT pipeline, sql_task, dbt_task — with its own parameters and its own compute.
Without this step
One notebook with 40 cells is a single task. It gets one duration in the run history, one retry granularity, and one row in the Runs table. When cell 31 fails you re-run all 40.
Four tasks exist. What decides the order they run in?
