TL;DR: Alex Monahan and Jacob Matson from the MotherDuck DevRel team walk through DuckDB 1.5 ("Variegata") live. They cover the new VARIANT type, the overhauled CLI, a built-in GEOMETRY type, faster joins, and improved concurrency, then demo the features and take audience questions.
What's in DuckDB 1.5
DuckDB 1.5 is a big release. The headlines are the VARIANT type for semi-structured data (think Snowflake's VARIANT, but in DuckDB), a redesigned command-line client with color themes and dynamic prompts, and a native GEOMETRY type that handles spatial data without an external library. Alex and Jacob go through each one, explaining what changed and why it matters.
VARIANT type and JSON performance
The VARIANT type stores typed binary data per row instead of raw JSON text. JSON analysis can be up to 100x faster than in previous versions. If you work with semi-structured data from APIs, logs, or event streams, this is the change you'll actually feel. Alex shows how the type works and where it fits alongside DuckDB's existing JSON and Parquet support.
Faster joins and better concurrency
DuckDB 1.5 can now use its hash join algorithm in more cases where it previously fell back to a slower nested loop approach. Complex joins against messy, real-world schemas get noticeably faster. Checkpoint concurrency is also improved — if you're running DuckDB with concurrent reads and writes, which comes up a lot in production MotherDuck deployments, that matters.
CLI overhaul
The new CLI has a fresh color scheme, dynamic prompts showing your current database and schema, a built-in pager for large result sets, and better customization options. If you spend time in the DuckDB terminal, the day-to-day experience is noticeably better.
DuckLake and lakehouse updates
DuckLake reaches v0.4 with macro support, sorted tables, and deletion inlining. The Delta Lake and Iceberg extensions also got updates, including write support improvements through Unity Catalog and AWS Glue. For more on how these formats fit together, see the open lakehouse stack overview.
Related Videos
7:14
2026-04-23
A Practical Guide to Context Management for Data Agents
Learn to write effective context for data agents. Covers benchmarks comparing English, SQL, and semantic model formats plus scaling strategies.
Webinar
AI, ML and LLMs
2026-04-21
MotherDuck Now Speaks Postgres: Fast Analytics Without Changing Your Stack
MotherDuck's Postgres endpoint lets any Postgres client, driver, or BI tool query your data warehouse without installing a DuckDB library.
Webinar
MotherDuck Features
Ecosystem
SQL
60:41
2026-04-09
Zero-Latency Analytics in Your Application with Dives
BI tools were never built to be app interfaces — they're rigid, clunky, and add complexity to your user experience. Dives offer a different approach: interactive data apps you create with natural language that can be embedded directly into your applications. In this session, Alex Monahan walks through how to build a Dive and embed it in your app, with help from AI agents. You'll learn how to create interactive visualizations with natural language queries, embed Dives into your app with a secure sandbox, set up the auth flow so your users get read-only access without exposing credentials, and choose between server-side and dual execution with DuckDB-Wasm. Whether you're building customer-facing analytics or internal tools, this webinar shows you the full workflow from query to production-ready embed. See how Claude Code + Dives can get you from zero to a working data app fast.
Webinar
MotherDuck Features
BI & Visualization
App Development