Near Earth.

Which near-Earth asteroid would a mining company visit first? These are all the asteroids NASA/JPL has found a round trip to, ranked by how cheap they are to reach, how big they are, what they are likely made of, and how many launch windows they have. Change the goal or the weights and the ranking re-sorts. How this works, in plain words.

Mining for

Reach by

NASA NHATS: leave low Earth orbit, stay 8+ days, come home.

What matters your weights

Filter

Max round-trip Δv: Minimum size Composition evidence

Cost to reach (Δv) vs size, one dot per asteroid. Top 10 are ringed. Click a dot for details.

Asteroids ranked by prospect score. Choose a name for its details.
Rank	Asteroid	Score	Δv	Size	Made of	Next launch	Trip

How this ranking works, and what it can't tell you

Data: NASA/JPL NHATS + Small-Body Database, NEOWISE years 4-7, MANOS and MITHNEOS spectral surveys · rebuilt daily · briefs by the R0ice Resource Team · How /neo works

Almost all of these asteroids have never had a spectrum taken, so their place in the ranking is partly a guess. This list shows the unmeasured ones where a single telescope measurement (a spectrum) would change the answer most: each could rise to the rank shown if it turns out to be the right kind of rock, and each can be observed in an upcoming window NASA has already computed.

Mining for

Unmeasured asteroids where one spectrum would change the ranking most
Asteroid	Rank now	If measured	Next window	Peak brightness	Telescope
Loading…

How this list is made

For each unmeasured asteroid we re-score it four times, once as each kind of rock (carbon-rich C, stony S, metal-rich X, basaltic V). Its size is recomputed each time, because an unmeasured asteroid's size is estimated from its brightness plus an assumed reflectivity, and a dark carbon-rich rock is about twice as wide as a bright stony one that looks equally bright. The same scoring as the Prospect tab is used, at its default weights.

"If measured" is the best rank it could reach. The next window and peak brightness come straight from NASA/JPL's NHATS data. Magnitude 20 or brighter is within reach of a 2–4 m telescope; 20–22 needs a large (8 m-class) one.

This says where a measurement would matter. It does not say what the rock is.

Can an AI tell what an asteroid is made of from the numbers NASA publishes? We hid the measured class of every near-Earth asteroid whose composition is known, showed each model only the catalogue numbers, and scored its guesses against a four-bin albedo lookup table.

Loading…

AI model scores against the albedo lookup and class frequencies (log-loss, lower is better)
Model	Names shown?	Answered	Model score	Albedo lookup	Class frequencies	Accuracy

Scores are log-loss: lower is better. Every model is compared with both baselines on exactly the asteroids it answered. Runs with fewer than 60 answers are shown but not counted in the headline.

Check it or beat it

Answer key (CSV) · Method and how to reproduce · every model's raw answers are in challenge/answers/.

Each model saw 20 asteroids per prompt with the class hidden and returned four probabilities (carbon-rich, stony, metal-rich, basaltic). Each ran twice, once with names hidden and once with names shown. Doing better with names shown suggests it remembered famous rocks instead of reasoning from the numbers.