The machinery behind the map
How anyone knows the sky changed
The short version: photograph everything, subtract yesterday, and be honest about what the leftover dots probably are.
A camera that photographs the whole sky, every few nights
On a mountain outside San Diego, the Zwicky Transient Facility (ZTF) sweeps the northern sky every two to three nights, each exposure covering an area 250 times the full Moon. In Chile, the new Vera C. Rubin Observatory has begun doing the same for the southern sky, deeper and faster — up to seven million change-alerts a night. Nobody is looking through an eyepiece. The telescope, the pipeline, and the first round of judgment are all machines, which is the only way to keep up with a sky this busy.
Finding one changed dot among a billion stars
The trick is called difference imaging, and it is almost insultingly simple: keep a pristine reference image of every patch of sky, and each night subtract it from the new exposure. Everything that didn't change — a billion stars, whole galaxies — cancels to black. What survives the subtraction is the news.
ZTF produces hundreds of thousands of these surviving dots per night. Most are mundane: known variable stars doing their usual thing, asteroids passing through, satellite glints, cosmic-ray hits on the detector. The interesting ones — a new light where the reference shows nothing, sitting on the outskirts of a galaxy — are the needles this site is built to hand you.
The light curve is the fingerprint
A single detection tells you almost nothing. What identifies an event is how its brightness changes night after night — the light curve. A classical nova spikes and drops in days. A core-collapse supernova rises for weeks and lingers. And a Type Ia supernova — a white dwarf detonating at a known critical mass — traces so repeatable an arc that astronomers use them as standard candles: measure how bright one appears, compare with how bright they intrinsically are, and you've measured the distance to its galaxy. Surveys of these explosions are how we learned the expansion of the universe is accelerating.
When a black hole catches a star
Occasionally a flare rises not from a dying star but from the exact center of a galaxy — where a supermassive black hole lives. If a star's orbit carries it too close, the black hole's gravity stretches it apart in an afternoon; the debris spirals in and glows for months. These tidal disruption events are rare — a given galaxy hosts one every ten to a hundred thousand years — but photograph enough galaxies every night and you'll catch a few in the act. The purple markers on the front page are the current candidates.
Why we say "probably"
Machine-learning classifiers read each alert — the image stamp on night one, the growing light curve after that — and assign probabilities. This site reports them as given: an event is "87% a Type Ia," not "a Type Ia." Early guesses get revised; some candidates dissolve into instrumental artifacts; the weird ones resist classification for weeks. None of that is failure. You are watching science mid-sentence, before the answer is known — which is precisely what makes a live sky worth reading.
Where the data comes from, tonight and tomorrow
Our events come from the public alert stream of ZTF, read through ALeRCE, one of the community brokers that filter and classify millions of nightly alerts for anyone to use. As the Rubin Observatory's decade-long survey ramps up its public broker streams, its alerts will join the map — same page, deeper sky, and a southern hemisphere's worth of new nights.