How /neo works

1 What near-Earth asteroids are

Asteroids are leftover rubble from when the planets formed, 4.6 billion years ago. Most circle the Sun between Mars and Jupiter. A small fraction have been nudged onto orbits that bring them close to Earth's path around the Sun: those are near-Earth asteroids.

NASA's Jet Propulsion Laboratory lists 42,735 of them. They range from house-sized boulders to mountains several kilometres across. Some are so easy to reach that getting there takes less rocket power than landing on the Moon.

That is why people talk about mining them: a rock you can reach cheaply, full of water or metal, could supply things that are expensive to launch from Earth.

2 The fuel bill (delta-v)

In space, what a trip costs is not distance, it is how much you have to change your speed. Speeding up to leave orbit, slowing down to match the asteroid, speeding up again to come home: add those changes together and you get a number called delta-v, measured in kilometres per second.

Think of delta-v as the fuel bill. Every extra kilometre per second means a lot more propellant, and propellant is most of what a rocket weighs. A lower number is a cheaper trip.

NASA's NHATS survey tries millions of real flight paths for each asteroid: leave low Earth orbit, fly there, stay at least 8 days, fly home, within 12 km/s and 450 days in total. 7,092 asteroids pass that test. Those are the ones /neo ranks.

/neo can also rank by a one-way robotic trip (no return leg), using JPL's Mission Design tool. That answers "what does it cost to get a prospector there", not "what does it cost to bring ore home".

3 Why composition is the bottleneck

Reaching an asteroid is only half the question. The other half is what it is made of, and for almost all of them nobody knows.

Asteroids are tiny, dark and far away. Most are only visible as a faint moving dot in a large telescope, and only for a few weeks when they swing close. Measuring what one is made of needs telescope time that is scarce.

So for most rocks, /neo can only say "probably stony" with a stated amount of doubt. The ranking shows that doubt openly. The "Worth observing" list goes further: it points to the unmeasured rocks where a single measurement would change the answer the most.

Water is harder still. The strongest direct sign of it is a dip in the infrared near 3 microns, and that has been measured on only a handful of reachable rocks.

4 Albedo and spectra: reading a rock from its light

We can't touch these rocks, so everything comes from the sunlight they reflect.

• Albedo is how much light a surface reflects. Fresh snow reflects most of it; coal almost none. Carbon-rich asteroids are dark (they reflect around 5% of sunlight), stony ones brighter. Albedo alone narrows it down, but a dark rock and a bright rock can still be the same kind.
• A spectrum splits that reflected light into colours, like a prism. Different minerals absorb different colours, leaving tell-tale dips. That is how astronomers sort asteroids into the classes above, and it is the real measurement /neo is missing for most rocks.
• Size is often a guess too. A small bright rock and a big dark one can look equally bright from Earth, so if the albedo is unknown, the size can be off by a factor of two.

Where a rock has only an albedo, /neo uses a simple statistical model trained on the measured rocks to estimate the chances of each class. That model was checked against 503 asteroids that were measured later, and its stated chances held up.

5 The AI test

Can an AI tell what an asteroid is made of from the numbers NASA publishes, without seeing a spectrum? We hid the measured class of every asteroid whose composition is known and asked AI models to guess from the catalogue numbers alone. A test is underway, and the results will be published with the answer key so anyone can check them.

6 Where every number comes from

/neo rebuilds its data every day from these public sources. Nothing is typed in by hand.

• NASA/JPL NHATSWhich asteroids a round trip can reach, the delta-v, trip length and launch windows. cneos.jpl.nasa.gov/nhats
• JPL Small-Body Database (SBDB)Orbits, sizes, albedos, spin periods and measured classes for every known asteroid. ssd.jpl.nasa.gov
• JPL Small-Body Mission DesignOne-way robotic trips and launch dates. ssd.jpl.nasa.gov/tools/mdesign
• JPL SentryNASA's impact-monitoring list. Most rocks on /neo are not on it. cneos.jpl.nasa.gov/sentry
• JPL HorizonsWhere each asteroid will be in the sky and how bright, for "Can I see it?". ssd.jpl.nasa.gov/horizons
• NASA NeoWsThis week's close flybys.
• MANOS surveyDevogèle et al. 2019, visible spectra of mission-accessible asteroids. doi:10.3847/1538-3881/ab43dd
• MITHNEOS surveyMarsset et al. 2022, near-infrared spectra of near-Earth asteroids. doi:10.3847/1538-3881/ac532f
• NEOWISE, years 4–7Albedos and sizes from NASA's infrared space telescope. Masiero et al. 2020 doi:10.3847/PSJ/ab7820 and 2021 doi:10.3847/PSJ/ac15fb
• Water (the 3-micron band)Ground-based: McGraw et al. 2022 doi:10.3847/PSJ/ac8ced and 2026 doi:10.3847/PSJ/ae3de0. Spacecraft: Ryugu, Kitazato et al. 2019 doi:10.1126/science.aav7432; Bennu, Hamilton et al. 2019 doi:10.1038/s41550-019-0722-2
• Gaia DR3Reflectance spectra from ESA's Gaia spacecraft, used only where our reading of them matched experts' classes. Gaia Collaboration 2023 doi:10.1051/0004-6361/202243791

On /neo, each number carries a small label saying whether it is straight from NASA, derived from NASA data, estimated by a model, or chosen by you (the weights).

7 What this site does not do

• No dollar values. "This asteroid is worth $X trillion" headlines multiply a rock's mass by today's price on Earth. That price would collapse long before anyone sold a tonne, so the number means nothing. /neo ranks by how reachable, how big and how useful a rock is likely to be.
• No impact scaremongering. Impact risk comes only from NASA's own Sentry list, shown with NASA's own scales. For most rocks the honest answer is that they are not on the list at all.
• No invented facts. The short written briefs are generated by volunteer computers, and a brief is only shown if every number in it appears in the NASA data it was given. It is also rejected if it contradicts that data, for example by calling a well-known orbit uncertain.
• No false certainty. An unmeasured rock is shown as a set of chances, never as a confident label.