🦆🤓 When Denis Hirn implemented DuckDB’s first recursive CTE operator in 2020, correctness determined the design: evaluate the non-recursive term once, then evaluate the recursive term until the next working table is empty. That established the right semantic contract, but reusable runtime state was scoped too narrowly. In the upcoming DuckDB v2.0 release, we assign those scopes explicitly. #DuckDB’s# recursive CTE engine now treats recursion as one long-lived computation: it retains eligible epoch-invariant state, chooses execution modes from exact frontier cardinalities and physical work, probes keyed state directly and gives `USING KEY ... UNION` changed-key semantics. Read more in this preview of the forthcoming v2.0 release:

More likely than not, you’ve been using AI tools such as Claude Code for day-to-day work. You may have noticed that when AI needs to look at a data file, it makes use of Python, writes a small script, runs it, and then reads the output. It’s an approach that works most of the time, but it’s slow and not always accurate. The duckdb-skills plugin gives Claude Code a growing number of skills that use the DuckDB CLI to read data files, run queries, convert formats, explore object storage, work with spatial data, search the documentation and recall earlier sessions. Here’s how you can get started with the new plugin:

New blog post: DuckDB Table Functions in Java 🦆☕ In a large organization, data is spread across many systems, including relational databases, document stores, message queues, data lakes, and cloud data warehouses. The layer that ties these together is often a JVM-based distributed query engine. Teams increasingly add #DuckDB# to these Java environments for fast single-node analytics. The DuckDB #Java# client can register table functions written in pure Java, exposing any Java-accessible data source as a SQL table. That turns @duckdb into a single-node query engine for heterogeneous joins across remote systems and local files, with no export step. Dive into the rest:

Quack, quack: Here’s your chance to meet fellow DuckDB devs in person! 🦆 Join us for the @DuckDB Boston meetup on September 3rd starting at 5:30 p.m. Enjoy some drinks and snacks, meet other DuckDB users, and learn from the technical talks (just announced) including: • @szarnyasg from @ducklabs_com 👨🏻‍🔬 • @gotdan from @BostonChildrens on unlocking healthcare data with DuckDB and open standards 🏥 • Richard Wesley from DuckLabs on windowing improvements in #DuckDB# v2.0 🪟 (venue still TBD, btw) The meetup takes place concurrently with the @VLDBconf, so it’s the perfect chance to meet fellow database researchers and practitioners from the Boston area and from all around the world. 🌏🌍🌎 Admission is free, but registration is required: Will we see you there? 🦆🦆

DuckDB Monthly #45# is out. Spotlight: Vladimir Gribanov, creator of the DuckDB MSSQL extension (native TDS, no ODBC). His bulk-load tuning took 38M rows into SQL Server from 933s to 96s. Plus DuckLabs joins AWS, MotherDuck acquires Tower, and v2.0-alpha lands.

What’s Next for DuckDB? 🦆🤔 Hannes Mühleisen, co-creator of DuckDB, was recently interviewed by Kris Jenkins of the Developer Voices podcast on every challenge, reward, twist, and turn of building a database as popular as DuckDB. From the early days of the product to the most recent acquisition of @ducklabs_com by @awscloud, Hannes – and what it means for the future of the open source DuckDB, DuckLake, and Quack projects*. It’s an interview as much as a story – an odyssey of what it means to build not only an open source product, but also a community, a company, and what success looks like in the day to day journey along the way. 📺 Watch the video episode (or read the transcript) here: 🎧 Or, listen wherever you get your podcasts. *Spoilers: DuckDB, DuckLake, Quack, and other open source projects in the DuckStack will remain open source and the project’s intellectual property, MIT license, and trademark are controlled by the DuckDB Foundation. The DuckLabs company is joining AWS, but the OSS projects remain OSS and controlled solely by the Foundation. 🦆 🦆

DuckDB 2.0 alpha reads Parquet and CSV from S3 2x to 3x faster, with zero query changes. 1.5: each worker downloads, waits, decodes, repeats. 2.0: a download pool fetches ahead, workers only decode. Same laptop, same query: 18.8 s to 7.7 s. On by default. Full numbers, plus recursive CTEs and VARIANT:

DuckDB 2.0 rewrote recursive CTEs. Walking a 20,000-commit git history: 1.8 to 16 s on 1.5.5, 0.10 s on 2.0, every run. 1.5: re-read the whole table every round. 2.0: read once, build a lookup, touch only the frontier's rows. Deep parent/child chains are where it shows.

When #DuckDB-Wasm# was launched in 2021, databases could not be persisted: everything lived in the Wasm heap and vanished when the tab closed. Keeping data meant serializing tables to Parquet, storing the bytes in IndexedDB, and re-registering them on the next page load. This was doable, but had to be handled at the application layer and was not offered out of the box by DuckDB-#Wasm#. Modern browsers now ship the Origin Private File System (OPFS), a per-origin, sandboxed file system with random-access reads and writes. DuckDB-Wasm can use it as a storage backend. This post by @carlo_piovesan and @GeertjanW shows you how:

Observability data shouldn’t be trapped behind one interface. Observe on Apache Iceberg writes telemetry to your S3 bucket as Snowflake-managed Iceberg tables, so tools like Spark, DuckDB, Trino and PyIceberg can query it directly through an open catalog. Dive in to learn more: