In the spring of 1858, a thirty-eight-year-old woman in London sat down with a stack of military death registers and a set of coloured inks. She had spent the previous two years in Scutari, watching soldiers die not from Russian bullets but from cholera, typhus, and dysentery in the British army's own hospitals. The numbers were damning. Nobody in Westminster wanted to read them.
So Florence Nightingale drew a rose.
What we now call her coxcomb diagram is a circle of twelve wedges, one for each month of the war, each wedge fanned into coloured layers showing the relative scale of preventable deaths, wounds, and other causes. It is a piece of statistics. It is also a piece of design. The blue petals - deaths from infection - balloon out of the page like a slow, accusing bloom. You cannot look at it and not understand. That was the point. Nightingale sent copies to Queen Victoria, to the War Office, to every member of Parliament she could reach. Sanitary reform followed.
The first great work of data visualisation art was a weapon.
The lineage
We tend to talk about data visualisation as if it were invented in the 1990s by a Tufte disciple with a copy of Adobe Illustrator. It was not. The instinct to turn numbers into shapes - to make information legible by making it beautiful - is one of the oldest moves in the human repertoire. The Babylonian star catalogues were data visualisation. The medieval zonal maps, dividing the world into climatic bands, were data visualisation. The Mercator projection of 1569 was an act of statistical compression that distorted the planet into something a sailor could use.
But the modern lineage - the line that runs from Nightingale to your GitHub contribution graph - really begins about thirty years earlier than her rose, in the offices of a French civil engineer named Charles Joseph Minard.
Minard's march
In 1869, the year before he died, Minard published a single sheet of paper now considered by many statisticians to be the most beautiful information graphic ever made. Edward Tufte called it "the best statistical graphic ever drawn." It is one image, drawn in two colours, depicting Napoleon's 1812 march on Moscow.
A thick beige band leaves Poland carrying 422,000 men. As it advances east, the band narrows. Each man lost is a sliver of width subtracted from the line. The band reaches Moscow shrunken to 100,000. Then the army turns back. The return march is drawn in black, narrowing further at every river crossing - the Berezina, the Dnieper - until what staggers home is a black thread of 10,000 men.
Beneath the route, Minard drew a temperature scale. The black band of the retreat is annotated, in neat engineer's script, with the falling temperatures: minus twenty, minus thirty Reaumur. You watch an army die from cold and starvation in a single image. Six variables - longitude, latitude, army size, direction, temperature, time - resolved into a shape so legible that a child can read it.
This is data visualisation as elegy. It is also data visualisation as art. The sheet has been reproduced and hung in galleries for a century and a half. People who do not know what a statistic is have stood in front of it and wept.
Tufte and the discipline of less
A hundred and fifty years later, in 1983, a Yale political scientist named Edward Tufte self-published a book called The Visual Display of Quantitative Information. He had to mortgage his house to print it. Within a decade it was the bible of the field.
Tufte's central argument was a moral one disguised as a design one. He coined the term chartjunk for the gratuitous decoration that infested the business graphics of the 1970s - the 3D pie charts, the cartoon men holding bars, the glossy fades. Each piece of decoration, he argued, was theft. It stole attention from the data. The job of the visualiser was to maximise what he called the data-ink ratio: the proportion of the ink on the page that actually carries information.
From this discipline came one of the most quietly influential ideas in twentieth-century design: the small multiple. A grid of identical small charts, varying only by one dimension, that lets the eye compare across many series at once. Weather data by city. Stock prices by company. Bird populations by year. The small multiple is what makes the GitHub contribution grid legible. Fifty-two columns, seven rows, one variable per cell. Tufte would have approved.
From public data to personal data
For most of its history, the data being visualised was public. Census records. Mortality tables. Trade flows. The viewer was a civil servant or a scientist, and the goal was policy or knowledge.
Around the year 2010, something shifted.
Aaron Koblin, a young artist working with the Aaron Draplin school of restraint, took the FAA's flight tracking data for a single day - around 140,000 flights criss-crossing North America - and rendered it as a still and a video. The image is unforgettable. The continent is dark. The flight paths are thin lines of light. They braid themselves into hubs at Atlanta and Chicago and Dallas, then unravel into the dark. You are looking at a system - air travel - drawn, for the first time, by itself. Koblin's Flight Patterns hangs in the permanent collection of MoMA.
A few years later, Giorgia Lupi and Stefanie Posavec began a year-long correspondence they called Dear Data. Each week they would pick a theme - mirrors looked into, doors held open, smiles received from strangers - track it for seven days, and visualise it on the back of a postcard. The postcards were posted across the Atlantic and back. The result, eventually published as a book, is data visualisation in its most intimate possible form. A week of one woman's anxieties, drawn as a constellation of coloured dots. A week of another's library habits, drawn as a forest of branching lines. The data was nobody's but theirs. The art was for everyone.
The quantified self as portraiture
By the late 2010s, the move from public data to personal data had escaped the art world entirely. Spotify Wrapped arrived in 2016. Strava heatmaps. Apple's activity rings. The Oura ring. The Whoop strap. Suddenly every consumer technology product had a visualisation layer, and every visualisation layer had the same implicit promise: here is a portrait of you, drawn from what you actually did.
This is a deeper thing than it first appears. For most of human history, a portrait was made by a painter who looked at you. Then, briefly, by a photographer. Now it is made by a sensor that watched you while you were not looking - your phone counting steps in your pocket, your laptop logging keystrokes, your music app remembering every late-night song. The portrait that emerges is more honest than any painted one, because you did not pose for it. You did not even know it was being drawn.
When Spotify Wrapped lands in your inbox in early December, you do not really need it to tell you what you listened to. You were there. You played the songs. What it gives you is the shape of the year - the genres, the hours, the unlikely repeat of a track you'd forgotten you loved in March. It is your year, drawn from outside. People share their Wrapped not because they need to brag about taste, but because the shape itself feels like a likeness.
The contribution graph belongs in this lineage. So does the running heatmap. So does the sleep chart. They are all small, honest portraits drawn by sensors that were already paying attention.
The atlas
Which brings us, eventually, to why we built Codex Stellarum.
There is a particular variety of data visualisation that resists the screen. The Minard map was meant for paper. Nightingale's rose was meant for paper. The medieval star atlases - the Uranometria, the Atlas Coelestis - were meant for paper, bound in vellum, opened on a desk under a candle. Some information wants to be held still. Some wants to be hung.
The GitHub contribution grid, in its native habitat, is a small green rectangle behind a username. It is honest data. It is also designed for a tab you close in thirty seconds. We started Codex Stellarum because we believed there was a richer object hiding inside it - the same data, treated with the dignity of a four-hundred-year-old star atlas. Our contribution charts are not screenshots of the green grid. They are redrawings of the same underlying data in the visual language of celestial cartography: stars sized by activity, constellations by week, the year's rhythm laid out the way Bayer laid out Orion in 1603.
We are, we hope, doing something Nightingale and Minard would have recognised. Taking a record that already existed. Rendering it in a form that asks to be looked at slowly. Letting the shape do the arguing.
Coda
Data visualisation is older than the computer, older than the printing press, older than the alphabet. It is one of the things humans do when they want to be understood. The best of it - the rose, the march, the Dear Data postcards, the small green grid that quietly logged the years of your working life - is also art, in the only sense of the word that matters: it makes you stop, it makes you see, and it does not let you forget.
If you would like to see your own data given that treatment, point us at your GitHub and we will draw your sky.