The Nearest Earth Nobody Has Seen

Proxima Centauri, a red dwarf 4.25 light-years away, moves toward us and then away from us at roughly the pace you use walking to the kitchen. That wobble — teased out of starlight by HARPS at La Silla and UVES at the VLT, across years of patient spectroscopy — is what a team of 31 scientists led by Guillem Anglada-Escudé announced on August 24, 2016, in a Nature paper dated the following day. Ten years ago this week. Something is tugging on the closest star to the Sun, on an 11.2-day cycle, with at least 1.27 Earth masses of tug. And it is orbiting at 0.05 AU, which around a star that dim puts it squarely in the temperate zone where liquid water is thermodynamically permitted.

The headlines wrote themselves. Earth-like planet found next door. The Planetary Society's coverage that day was more careful than most, and even it reached for the goldilocks framing.

Here is what ten years bought us.

The mass estimate got sharper: 1.17 ± 0.09 Earth masses, after ESPRESSO independently confirmed the signal in 2020. The period got absurdly precise: 11.18465 ± 0.00053 days, which is to say we know the length of Proxima b's year to about forty-five seconds. The tilt got argued about — the star's rotation axis leans roughly 47 degrees to our line of sight, and if the planet's orbit is coplanar with it, the true mass lands near 1.44 Earth masses, comfortably rocky-sized. Nobody has measured the radius. Rocky is an inference about a size class, not an observation of a rock. The neighborhood filled in — Proxima d confirmed in 2022, orbiting even closer; Proxima c reported in 2020 and since disputed as a possible artifact.

And we still do not know whether it has air.

Not "the data are ambiguous." Not "results forthcoming." We do not know, and the reason is not that anyone has been lazy. It's geometry. Proxima b does not transit. From where we happen to be standing in the galaxy, the planet never crosses the face of its star — the orbit is tilted the wrong way, and it will stay tilted the wrong way for as long as anything we build will last. Transit spectroscopy is the workhorse method for reading an atmosphere: wait for the planet to pass in front, catch the sliver of starlight that filters through its air on the way past, look for the fingerprints of carbon dioxide and water and methane. It is how JWST does nearly all of its atmospheric work. It requires the planet to get between you and the lamp. This one refuses.

And that breaks the intuition the word nearest was quietly carrying. Proxima Centauri is the closest star there is; there will never be a closer one, not with better funding and not with a better century. You would expect that to buy something. It buys remarkably little. About forty light-years out — nearly ten times farther — sits TRAPPIST-1, and JWST has real constraints on the atmospheres of its inner planets, largely negative ones, because those planets happen to cross their star. Distance lost to geometry, and it was not close.

So the difficulty curve for characterization is not ordered by distance at all. It is ordered by alignment — whether a planet's orbital plane happens to point edge-on at Earth, a coin flip settled billions of years ago when the system condensed and unappealable ever since. We sit 4.25 light-years from a temperate rocky world and cannot read its air, while we read one forty light-years out. Proxima b is not the floor of the difficulty curve. It is off the curve entirely, behind a wall that has nothing to do with how far away it is.

We have something north of six thousand confirmed exoplanets now. That number is quoted like an accomplishment, and it is one, but it should be read for what it is: a census, not an acquaintance. Detection scales beautifully, because detection is a statistical operation performed on starlight you already collected. Characterization barely scales — only across the geometrically cooperative subset, and inside that subset at a punishing cost per target, because characterization needs photons that touched the planet, and those photons are few, faint, and drowned in a glare ten billion times brighter sitting a hairsbreadth away in the sky. JWST has done real atmospheric work on dozens of worlds. It will never do any on this one.

Nobody has ever seen Proxima b. Not once. Not a pixel. Everything above is inference from a star's slow lean. The proposed fix has been the same for ten years — build 30-meter-class telescopes with extreme adaptive optics, ELT and TMT and GMT, and pull the planet's own light out from under its star's — and those are still, in 2026, mostly under construction.

Meanwhile the star flares. Proxima throws sporadic UV and X-ray tantrums that could have stripped the planet's atmosphere billions of years ago, or sterilized whatever was standing on it, or done neither because a strong enough magnetic field or a thick enough envelope shrugged it off. All three remain live.

The rotation is open too. A planet orbiting that close should have had its spin locked by tides long ago, which would mean a world 4.25 light-years away with a year eleven days long and the same face pointed at that red star for four billion years — one hemisphere in permanent noon, the other in permanent night, a ring of eternal sunset between them where the interesting weather would be. Or it settled into a resonance instead and has proper mornings. Both are still on the table.

Now the part that makes the blank worse. Four months before the discovery, in April 2016, Yuri Milner announced Breakthrough Starshot — a hundred million dollars toward gram-scale probes riding laser sails to a nearby star at a fifth of light speed. When Proxima b turned up that August it became the destination, and it has been the destination in every serious interstellar-probe sketch since, for the unromantic reason that it is the only exoplanet a human artifact could plausibly ever reach. Starshot itself has gone quiet — under ten million of the hundred spent, no prototypes, no announcement of an ending either. The program faded. The target did not, because there is no other one.

So this is not the ordinary gap between a catalog and an acquaintance. We have seriously proposed sending something to a world we cannot establish has air, and the geometry blocking that answer will still be blocking it on the day anyone has to decide whether to launch.

That is not ignorance, exactly. Ignorance is cheap and comes free. What we built over ten years is bounded ignorance — we know precisely what we don't know, precisely why, and precisely what instrument would fix it. That's expensive, and it's rarer than a telescope.

We know the address. We have never met the occupant.