| | If models or dashboards break when upstream data changes, APIs are often the culprit. This live workshop shows how Protobuf helps teams evolve schemas safely, avoid breaking changes, and enforce data quality across services and streams without turning API design into a bottleneck.
*Message from this week's sponsor, Buf |
| | |
| < | | | What if a model’s prediction was literally an eigenvalue? This post kicks off a short series on spectrum-based models, exploring interpretability, robustness, and scaling. Start here, then follow the links to go deeper.
Alex Shtoff |
| | |
| | Here's a practical guide to speeding up NumPy with measurable wins. Learn when threading helps, when Numba helps more, and why memory bandwidth caps your gains. Ends with a clear-eyed take on why “automatic” parallelism can be risky in real code.
Python⇒Speed | Itamar Turner-Trauring |
| | |
| | If you’ve got piles of text like support tickets, survey responses, or abstracts and need structure fast, this is a solid playbook. Covers cleaning data, choosing models, tuning parameters, merging topics, and sanity-checking results with visuals and qualitative evaluation.
Institut Polytechnique de Paris | A. Morin, É. Schultz, É. Ollion |
| | |
| | Ever add an index and wonder why things didn't get faster? This clear, hands-on guide explains how PostgreSQL actually stores data, when indexes help, and when they quietly make things worse. Lots of real EXPLAIN output, real tradeoffs, and zero hand-waving.
Dalto Curvelano |
| | |
| | | | Not all data is meant to become a dashboard. This new project curates striking works that turn datasets into art, blending scientific structure with abstraction and emotion.
Yan Holtz and Marthe Viallet |
| | |
| | | This new R package shows how far database-backed analytics have come. dbreg runs linear models directly on database backends, shrinking huge tables into tiny regression-ready objects. Inspired by the duckreg Python package, it pairs familiar R ergonomics with DuckDB support out of the box.
GitHub | Grant McDermott |
| | |
| | msgvault is a local-first email archive that treats your inbox like an analytical dataset. It syncs millions of emails into SQLite, builds Parquet indexes, and uses DuckDB for millisecond search, complete with a fast TUI, CLI, and MCP server for local LLM workflows. It's alpha, but packed with great ideas.
Wes McKinney |
| | |
| | | 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. |
| | |
| | | A Facebook for AI agents? Moltbook is a weird, fascinating experiment where bots install themselves via markdown, talk to each other, and share real automation tricks. This isn't about data directly, but worth paying attention to as agent ecosystems start taking shape.
Simon Willison |
| | |
| | | This week's issue sponsored by: | | |
|
|
