dbt Core vs StatsBase.jl

dbt is running two release lines at once. The 1.x Python line reached 1.12.0 in July, a GA that removed the experimental dbt login command and the bundled dbt-state plugin outright while adding the v2 semantic layer YAML parsing. The 2.0 Fusion line — the Rust engine — moved through five alphas and reached its first beta on August 10, carrying catalog-free binding, dbt state explain, redundant-test skipping, and a lint rule system that understands node selection.

Fusion is being built to do statically what dbt-core did by asking the warehouse. Catalog-free binding lets SQL bind without introspection, tests get skipped when they are provably redundant, and dbt State speculatively submits nodes while the dependency prefetch is still in flight — all of it trading round-trips for compile-time analysis. Meanwhile 1.x is absorbing the v2 semantic layer YAML piece by piece, which puts metrics and entities into the model graph itself. Adapter breadth keeps widening in parallel, with Databricks service principal auth, Redshift group grants, and ClickHouse materialization configs.

With beta.1 out, the next milestones are further betas hardening the Fusion feature set toward parity, and continued v2 semantic YAML work landing in the 1.x line.