The blog is open.
Long-form notes on data engineering, AI experiments, and the seam-work of building systems that actually ship. New posts when I have something worth saying.
DuckLake is the new Delta Lake
Same open-lakehouse job Delta made mainstream. Different catalog (SQL tables, not metadata files). Real governance hole. A working claim with the limits left in.
Replace Your SQL Pipelines with Polars
Your transformation stack is four tools and three languages. A Polars, Dagster, and pytest stack is one language, any data source, and data contracts enforceable in CI.
SQL or Python for Data Transformations? Start With the Engine
SQL won the transformation layer. Python won the machinery around it. A field note on choosing by execution engine, data location, logic shape, and maintenance discipline.
Your Databricks Lakehouse Is a Postgres Table and a Parquet Folder
A working note for anyone who signs a cloud bill. DuckLake is an open lakehouse format that stores metadata in a real SQL database and data in Parquet. The bill when you put it on a European box is 80–90% lower than the bundled alternative.
A Hetzner Box vs. a Databricks cluster
A version of a working argument, written for anyone who has ever looked at a cloud bill and wondered what they are paying for. The short version: most companies are renting a fleet of trucks to deliver an envelope.
Three CEOs on what work becomes
Sam Altman, Dario Amodei, and Elon Musk share a direction. They split on the speed, the tone, and what to do about it. A research note on the divergence, with the data underneath.
Why Multi-Agent Systems Work
A single large model, asked to do four jobs at once, will blend them. The fix is structural — not a bigger model, but a smaller contract. A field note on separation, handoffs, and the hard part of multi-agent systems.