dbt Core vs Apache Iceberg

dbt-core is releasing on two tracks at once. The Python line reached 1.12.0 on 16 July after three release candidates, and it is a tightening release: the experimental `dbt login` command and the bundled dbt-state plugin were removed outright, and flags introduced in 1.9 and 1.10 now default to true. The 2.0.0 alpha track is the Fusion engine, and its work is almost entirely about catalogs — read-write Horizon and Unity access over Iceberg REST via DuckDB, a catalogs.yml v2 covering DuckLake, Iceberg REST and local filesystem, plus catalog_database overrides and Redshift catalog generation through SHOW TABLES and SVV_REDSHIFT_COLUMNS.

The division of labour between the two tracks is clear from the entries: 1.x is consolidating and removing experiments, while 2.0 is where the new surface area lands. The 2.0 surface is specifically the lakehouse catalog layer — dbt is moving from a tool that writes to a warehouse toward one that binds to open table catalogs directly, with materialization made catalog-aware. Notably 1.12.0rc1 also teaches the Python engine to tolerate Fusion-specific warn_error_options rather than erroring, so the two engines are being made to coexist in the same projects rather than fork.

The alphas are still expanding catalog coverage adapter by adapter, so expect further catalog integrations and continued catalogs.yml v2 work before 2.0 leaves alpha. On the Python side, with the deprecated flags now defaulted and the experimental commands removed, 1.12 looks like a stabilization point rather than a base for new features.