Tools & Tutorials
Layers: "How we packed our hand plotted dot art into a very compact binary format, reduced geometry delivery by 93% and stopped rendering the background entirely when nothing was changing."
They did this by creating a custom binary format that shrank 103,802 hand-placed dots from 1.25 MB to 82 KB. Splitting data into separate planes drove most of the savings.

"This wiki sets out four steps to take before you submit a paper, so that anyone who downloads your code can run it and get your results: make your code Transferable, Available, Documented and Annotated."
It offers a pre-submission checklist plus guidance on README/licence writing, repository choice, and code review before paper submission.
Onibaku.Apps has launched tools that use UK Government Open Data: "planning constraints, food hygiene ratings, the Charity Commission register and Ofsted. None of them scrape anything. Each time you run one, it reads the official API or open data file there and then, and gives you back one flat row per thing, so you're never looking at an old copy."
Geospatial foundation models (GeoFMs) learn general representations from Earth observation data, producing embeddings reusable across tasks like classification, retrieval, and change detection.
"Embeddings are one way we make use of the rich representations learned by geospatial foundation models, supporting everything from search and retrieval to natural language interfaces. Yet for many practitioners, the concept remains opaque. This primer explores how embeddings are generated, the kinds of questions they help answer, and why they’re useful across the geospatial ecosystem."

The Benchouse team reports how switching from Sonnet 5 to Opus 5.5 lifted accuracy from 49.6% to 88.2% using only bare schema, before any context work, and illustrates how adding full documentation, context-rule writing, and explicit business definitions pushed Opus to 96.5% accuracy at 100% completeness.

"By “dynamic” social media cards, I mean cards for pages of user-generated content or something like an e-commerce site with loads of different products. Meaning too many pages to reasonably design a custom graphic by hand for each. So what we’re talking about here is a system to generate dynamic images with dynamic data."
Different platforms (Discord, Slack, X, iMessage, etc.) render "unfurls" (apparently this is the term...) very differently, so you can't fully control how a card appears.
