SPACESIFT
What can telescopes find, and what do they miss?
Ninety days of a star's brightness. A planet is hiding in here.
Move your cursor: it bends starlight the way gravity does.
Three investigations, one question: how often is a search right, and what does it miss?
Every result shows its data, method and baseline, counts its errors in both directions, names its limits, and can be rerun.
Every night, telescopes record millions of points of light. Most are noise. Some are new worlds. A few could come close to Earth. SpaceSift asks how often we catch them, and how often we are fooled.
Worlds hide in starlight.
When a planet crosses in front of its star, the starlight dips. For an Earth passing a Sun-like star, by less than 0.01%.
Searches hunt for the repeat.
Planet-hunting software scans months of brightness for a dip that comes back, orbit after orbit.
But stars flicker.
Pulsations and starspots can drown the signal. In our tests the standard search found 0 of 106 planets hidden around a pulsating star, and 105 once its rhythm was removed.
Hide a planet yourself →Can you find the planet before the computer does?
A planet is hidden in this star's light. Slide the orbit until its dips stack into one: the signal strength jumps when you hit it. The trace under the slider remembers every orbit you've tried.
Fine-tune with − and + (or Shift + arrow keys). A green spike in the trace is worth going back to.
Light is a fingerprint.
Spread starlight into a rainbow and dark lines appear: atoms in the star soaking up exact colors.
Every line names an element.
Calcium, hydrogen, sodium, iron. Their pattern reveals how hot a star is and what it is made of, and the same trick tells rock from metal on an asteroid.
Read any star.
Upload a spectrum, even a picture of one, and it is matched against 123 real stars. Or test yourself on a mystery star.
Read a star →New dots, every night.
Survey telescopes report moving objects they can't identify to the Minor Planet Center's confirmation page.
Which ones are real?
A machine-learned ranking scores every candidate twice a day by how likely it is to be a genuine near-Earth object, not junk or an ordinary asteroid. That is not the chance of hitting Earth.
Catch them before they're gone.
A candidate nobody follows up can be lost within days. Ranking them is a small, real piece of planetary defense.
Open the live dashboard →Every new asteroid candidate, where it is right now.
The sky drawn as a globe, with Earth at the centre. Each dot is a candidate on the Minor Planet Center's confirmation page, glowing by how likely our model thinks it is to be a real near-Earth object. Drag to turn the sky.
Loading tonight's candidates…
- Likely real (P ≥ 0.8)
- Uncertain (0.4 to 0.8)
- Probably not (under 0.4)
- Ecliptic, where the planets and most asteroids travel
- Celestial equator
A likely-real candidate far from the ecliptic is especially interesting: ordinary main-belt asteroids stay close to it. A high score means "probably a real near-Earth object", not "on course for Earth".
Open the full ranking →Small signals, measured honestly.
How much an Earth dims a Sun-like star as it passes in front. That is the signal planet searches have to find under noise and stellar flicker.
Planets hidden around a star pulsing every 4 hours (experiment SS-0002A).
A library from hot blue O stars to cool red giants that uploaded spectra are matched against.
Every new candidate on the Minor Planet Center's confirmation page is re-scored.
WHAT IT IS
Three free tools and the open experiments behind them: a lab for hiding planets and testing the search, a lab for reading what stars and asteroids are made of, and a live ranking of new asteroid candidates.
WHO IT'S FOR
Students, observers and anyone curious how we know what's out there, not just the headline results. Every number comes from code and data you can rerun.
WHY IT MATTERS
A missed planet looks exactly like no planet, so counts of how common worlds are depend on knowing what searches miss. And a new asteroid that isn't followed up in time can be lost.