| | | If you’re like us, you got into tech because you’re driven by the eternal question: “what if?” That curiosity is woven into the Fidelity workday. We want our technologists to not only stay informed, but to continue sharpening their skills and become the best at what they do. Here, we build both cutting-edge products and cutting-edge careers. Read more about developing skills at Fidelity.
*Message from this week's sponsor, Fidelity Careers |
|
| | 🔗 Note: if you don’t see links in this email, it’s also available on the web: dataelixir.com/archive/issue-586.html |
| | |
| | Given a single photo of an unknown island resort, the challenge here is to identify the resort, its coordinates, and the camera direction. Reverse search tools and AI aren’t allowed. Good luck.
Yassa Sfen |
| | |
| | This is a thoughtful workflow for getting better results out of AI from someone who's been using it at scale inside Google. The post is geared towards engineering but the ideas apply to analysis, data pipelines, modeling, and other work that's common on data teams.
Dave Rensin |
| | |
| | A watermark for AI text doesn’t need to be hidden in metadata. It can be embedded in the statistical pattern of the words themselves. This interactive explainer shows how that works and why the signal gets harder to detect as text is rewritten.
James Padolsey |
| | |
| | This post explores a supply-chain model that seemed to be 97% accurate before information was removed that it wouldn’t actually have at prediction time. It's a great example of how data leakage can turn an impressive benchmark into something much less meaningful.
GitHub | Ilpo |
| | |
| | A Tier 1 bank benchmarked AI queries with and without a semantic layer and found a 21,000x efficiency gain. See the production benchmark, the SQL behind it, and why prompt engineering, catalogs, dbt, and MCP alone couldn't close the gap.
// sponsored |
| | |
| | | | Aman Bhargava wanted to map cloud cover across India, but the data was trapped inside hundreds of weather charts. So he read the pixels back out of the charts, reconstructed the underlying data, and turned it into an interactive map of the sky. Great project!
Aman Bhargava |
| | |
| | | Polars and SQL can express many of the same operations, but they don’t always mean the same thing. In this post, Marco Gorelli walks through some of the subtle differences that can trip you up when moving between the two.
Marco Gorelli |
| | |
| | The upcoming DuckDB 2.0 is a major new release. This version adds server mode, support for semi-structured data, new SQL features, faster queries, and more.
Mark Raasveldt and Hannes Mühleisen |
| | |
| | | This is a great resource for experienced Python coders who want to write clearer, more idiomatic Python. Covers a wide range of topics from type hints and testing to generators, concurrency, NumPy, Numba, metaprogramming, and performance optimization.
Bruce Eckel |
| | |
| | | You can learn a lot about optimization by watching it fail. This interactive post lets you manipulate a genetic algorithm as it evolves an image, so you can see how mutation, selection pressure, diversity, and convergence affect the solution.
Ken Reid |
| | |
| | | This week's issue sponsored by: | | |
|
