| | ETL and ELT are never “set and forget.” This free webinar walks through real patterns for building pipelines that adapt as data changes, covering ETL vs. ELT tradeoffs, dynamic DAGs, and what Airflow 3 unlocks in practice. Register for the free webinar.
*Message from this week's sponsor, Astronomer |
| | |
| Note: Some readers have reported that Exchange/Outlook is stripping out the links in Data Elixir. I’m told it will eventually fix itself but in the meantime, if you don’t see links in this email, it’s also available on the web: dataelixir.com/archive/issue-560.html |
| | |
| | | If residual plots could lie to you, would you notice? This post shows why transforming covariates doesn’t change your outcome’s distribution, how mis-specified models distort residuals, and why choosing OLS vs something else requires real thought about the data-generating process.
Robert Kubinec |
| | |
| | How do you make AI actually useful for day-to-day analytics without breaking trust? This behind-the-scenes look at OpenAI’s internal data agent shows what it takes to go from natural-language questions to reliable answers across messy, real production data.
OpenAI Engineering Blog | Bonnie Xu, Aravind Suresh, and Emma Tang |
| | |
| | Pretraining works great for images and text, but tabular data is messy. This post breaks down how models like TabPFN pull it off, using Bayesian intuition, synthetic task priors, and in-context learning to make tabular foundation models actually transferable.
Mindful Modeler | Christoph Molnar |
| | |
| | Not all code is meant to live forever. This post makes the case that LLM-generated throwaway code unlocks new ways to extract, enrich, and explore data that were previously too expensive to bother with.
Counting Stuff | Randy Au |
| | |
| | In this deep dive, Simon Späti explores what happens when you treat a folder of notes like real data. The result is a thoughtful look at how structure and semantic search can surface forgotten work, related ideas, and unexpected connections across your own writing.
MotherDuck | Simon Späti |
| | |
| | | | This goes way beyond a single course. Alberto Cairo’s Open Visualization Academy is shaping up to be a full, free curriculum in data visualization and information design, with multi-hour courses on perception, ethics, accessibility, storytelling, and chart mechanics. Taught by instructors from across the visualization community.
Alberto Cairo |
| | |
| | This new project takes natural language as an input and then searches for related visualizations from TidyTuesday contributions, Datawrapper's Data Viz Dispatch, and FlowingData. Works for finding examples of specific types of visualizations, color encodings, chart elements, statistical methods, dataset types, and more.
My Next Chart | Muhammad Chenariyan Nakhaee |
| | |
| | | One year, fifty releases, countless design decisions. This behind-the-scenes look at building data quality tooling highlights how APIs, ergonomics, and reliability evolve over time, and why maintaining trust in data systems is harder than it first appears.
Posit | Rich Iannone |
| | |
| | | In addition to Data Elixir’s regular weekly edition, we’re experimenting with community-curated, focused editions built from Bluesky posts. These surface high-signal links, with direct access to the surrounding discussion so you can read, follow, and join in when you want. | We’re starting with two editions: | Data Practice — Everyday data work, including SQL, statistics, visualization, R/Python, and tools. Published Wednesdays. Learning Machines — Machine learning and AI, including tutorials, research, tools, and applications. Published Thursdays.
| Opt in for free. |
| | |
| | | This week's issue sponsored by: | | |
|
