There's a vast amount of strange, freely-downloadable data on the internet — paintings of horses, sightings of frogs, every earthquake in the past hour, scans of sixteenth-century maps. This is a list of the ones I keep coming back to, plus a few I've taken apart and visualised.
Visualisations
Every recorded meteorite landing on Earth~32,000 dots from NASA's catalogue, plotted by where they fell. The continents draw themselves — but the densest clusters aren't where you'd expect.
Atlantic hurricanes since 1851All 1,973 storms in NOAA's HURDAT2 archive. The naming convention is younger than you think, and the busiest decade isn't the one you'd guess.
Every Nobel laureate, plotted by age at award1,026 prizes since 1901, scattered by year and age. The trend nobody noticed for fifty years: the average laureate is over twenty years older than the first generation.
Eurovision, every entry since 1956Sixty-nine years of European pop in one stacked chart. Watch the languages of performance flip every time the contest changes its rules.
Museums That Opened Their Vaults
Met Museum Open Accessmetmuseum.orgEvery painting, sculpture, costume, and weapon the Met owns — half a million objects dropped into the public domain in 2017. You can grep the whole collection.
Smithsonian Open Accesssi.eduWhen a 175-year-old institution releases 4.5 million items as CC0, you get fossils next to spacecraft components next to portraits, all in one folder. This is that folder.
Rijksmuseum APIdata.rijksmuseum.nlVermeer, Rembrandt, and 700,000 other Dutch artifacts at zoom-deep resolution. Don't expect to leave the site quickly once you start scrolling.
MoMA Collectiongithub.com/MuseumofModernArtMoMA put their entire artwork and artist database on GitHub. With version history. You can git pull modern art.
Earth, Sky, and What's Below
CERN Open Dataopendata.cern.chActual collision events from the Large Hadron Collider — the same files physicists publish from. It's a few terabytes and it will not run on your laptop.
GBIFgbif.orgEvery documented sighting of every catalogued organism, ever. 2.6 billion records. Where biologists go when they need to know if a species has shown up somewhere new.
USGS Earthquake Feedearthquake.usgs.govA live JSON feed of every earthquake on Earth, updated within a minute of detection. Plug it into anything that needs to react to the planet moving.
NASA Open Data Portaldata.nasa.govEvery NASA mission since forever, dumped into one searchable portal. Mars rover imagery, exoplanet catalogs, decades of climate-satellite data.
The Web, In Bulk
Common Crawlcommoncrawl.orgMost of the open web, re-downloaded monthly, hosted on S3 for anyone to use. This is what your favourite language model started life eating.
Wikidatawikidata.orgWikipedia's neglected sibling — a structured database of every entity in Wikipedia, queryable in SPARQL, downloadable in full. Once you start using it you wonder how you searched anything before.
Internet Archivearchive.orgThe Wayback Machine is just the front of the shop. Behind it: the largest privately-funded library on the planet — books, audio, video, software, web pages.
Project Gutenberggutenberg.org70,000+ public-domain books as plain text. The grandfather of all open-data projects, still going since 1971.
People, Places, and Public Records
GDELTgdeltproject.orgEvery news article published anywhere in the world, tagged by who, where, and what tone — refreshed every fifteen minutes. Try not to fall in.
OpenCorporatesopencorporates.comIf a company has a registration number somewhere on Earth, it's probably in here. Useful for journalists; quietly alarming for anyone who searches their own name.
Our World in Dataourworldindata.orgLong-run human-development data done properly — every chart sourced, every series downloadable, every claim checked. The bar this site sets is the bar.
OpenStreetMap Planetplanet.openstreetmap.orgThe entire vector map of Earth in one 150 GB file. Every road, every building, every park bench. You can host your own Google Maps from this.
Odd, Wonderful, Specific
Million Song Datasetmillionsongdataset.comA million tracks' worth of audio features — tempo, key, danceability, mood. No audio, just the numbers. A lot of music-information-retrieval research starts here.
The Pudding Datagithub.com/the-puddingSource datasets for The Pudding's visual essays. Hip-hop vocabularies, gendered film dialogue, the shape of a typical horror movie. Each one already comes with a story attached.
Our Airportsourairports.comEvery airport, heliport, and seaplane base on Earth in a single CSV. Hand-maintained, kept current. A model of what an open dataset should look like.
David Rumsey Map Collectiondavidrumsey.com100,000 historic maps, scanned at the kind of resolution where you can read the engraver's signature. Pre-modern cartography as an internet rabbit hole.