Consider what a data engineer has to learn before writing a single useful transformation on a typical 2026 data stack.
dbt, for SQL models, sources, tests and macros. Airflow or Dagster, for scheduling, sensors and retry logic. dlt, Airbyte or Fivetran, for ingestion. Python scripts for anything SQL cannot express. Then YAML to configure all of it, and a growing pile of glue to hold the pieces together.
Each of those tools is good. dbt genuinely standardised SQL transformation. Airflow's operator ecosystem is unmatched. dlt does schema inference and incremental loading better than most hand-written extractors. The problem is not the tools.
The problem is the seams.
Where failures actually live
