Skip to content

News

The Old Shop's 99 Datasets, Pointed at Where the Data Lives Today

By glitchdata Team 2 min read

The old glitchdata shop sold 99 compiled datasets — postcodes, electoral boundaries, company registers, transport timetables, and a long tail of reference tables that were tedious to assemble in 2014 and are still tedious now. They have all been migrated to the hub.

What Migrating Actually Meant

Each product description was a page of prose with the useful parts buried in it. The importer pulled them apart into the fields a dataset page expects: the introduction, the metadata field list, the sources, the formats and the licence. Those are now structured, searchable and filterable, instead of being a paragraph somebody had to read.

Most of Them Have a Better Source Than Us

The harder question was honesty. Most of these were compiled by Glitchdata Research years ago, and for the great majority the same data is published today by the organisation that owns it — an electoral commission, a statistics agency, a standards body, a government open-data portal.

So each dataset records where the data can be had now, and that link is what the page leads with. If an official register publishes it, you should get it from the register: it will be current, it will keep being current, and it will carry the publisher's own licence.

Eight have no authoritative source left — the body that published them has gone, or the series stopped. Those still point at the compiled file, and the page says that is what it is rather than dressing it up.

The Hub Does Not Hold the Files

This is the same rule as everything else here: glitchdata references, it does not store. A dataset page catalogues what exists, previews the first rows of a table where it can, and counts downloads — then sends you to whoever publishes the file. Nothing is mirrored, so nothing here can quietly go stale while looking current.

More news

News 3 min read

Your Account Is Yours, and So Are the Numbers

A new Insights page reports your downloads, likes, progress and quiz results, with CSV exports of all of it. Alongside it: account settings, a forgotten-password link that actually exists, and a notification when an admin reviews your work.

News 2 min read

One Search Box, and a JSON API for Everything Public

Search now covers datasets, models, courses, guides and news from a single box. Everything a signed-out visitor can read is also available as JSON, with no key and nothing to sign up for — plus RSS feeds and a sitemap.

News 2 min read

Quizzes, Certificates and Reviews for Every Course

Lessons can now end with a quiz you have to pass to finish them, courses issue a shareable certificate when you reach the end, and the learners who took a course can say what it was worth.

Guide 2 min read

Five Hundred Guides, From First Principles to Agent Harnesses

The guide library has grown to 500 articles across sixteen topics — machine learning and statistics at one end, RAG, prompting, MCP, A2A and AI security at the other. Every one is free, needs no account, and sits one click from the courses that recommend it.