A new paper says the dichotomy is an illusion of detection. Daniel Yahalomi (MIT) and coauthors ran a planet-population model (ExoMOD) against Kepler's DR25 catalog of 2,557 candidate planets around Sun-like stars. Their model has no false positives, yet it still reproduces the faint one-blink excess.

In it, 85% of detected single-transit planets sit in systems holding more than one planet; Kepler missed the siblings. Past 100-day orbits, still 73%.

Why: spotting several planets around one star favors compact, short-period systems, which transit more often and dip deeper. Long-period planets give fainter signals, and their unseen companions land in the “single” pile.

Caveats: not peer reviewed (submitted to AAS journals, Oct 15 2026). The 85% is model output, not a detection; it assumes one shared parent population. The data settle the pattern; how many hidden planets exist is inference.

Source: arXiv 2610.22349v1, Yahalomi et al. (section VII).

A new paper gives a duller cause: star spots. Refractory elements tend to have low first-ionization potentials and strong lines, which makes them the most sensitive to the cool, magnetized air inside a spot. So a spotted star looks refractory-poor in a way that tracks Tc, with no planet involved.

Casey, Bedell and co-authors (Flatiron) show it three ways. More active stars show steeper Tc slopes, the least active near zero, across 79 Sun-like stars with activity re-measured from their own HARPS spectra. A simulated spotted Sun reproduces the reported slopes. And a sunspot umbra lifts many elements by up to a factor of three over the quiet Sun.

Caveats: preprint (Oct 13), not peer reviewed; the authors call their own fit physically unmotivated and warn against causal claims drawn from it; the simulations skip non-LTE, magnetic fields and molecular lines; and the umbra is an extreme, not a whole star.

This does not kill planet formation. It asks that any Tc-slope claim show a quiet star first.

Brinkman and Cowan (McGill) checked them against their stars. Of 333 small planets with mass and radius, four are rocky above the limit (WASP-84 c, TOI-1347 b, TOI-2322 c, HD 63433 b), plus four dense, volatile-rich worlds.

All eight orbit magnetically active stars, log R'HK above -4.75 (the Vaughan-Preston gap); 13 of their 37 stars clear it. Spread evenly, the odds all four rocky ones land on active hosts are about 1%. That should not happen: building one takes time, so they favor older stars.

Likeliest answer: inflated masses. Starspots mimic a planet's tug in radial-velocity data, and the fit can absorb or inflate the signal, especially near the star's rotation.

Caveats: preprint submitted to ApJL, a statistical argument over eight planets, not a re-fit. The paper's 0.03% for all eight doesn't follow from its own formula (0.003% carried out to eight terms, base unshown); I lean on the four. The volatile-rich ones are dense Neptunes, so “Mega-Earth” is fuzzy.

Source: Brinkman & Cowan, arXiv 2610.12600v1, submitted to ApJL, Oct 8 2026.

The record splits in two. The mineral surface holds: near 560 K, the same two minerals (1-micron olivine, 0.1-micron forsterite) for 17 years. The brightness does not. The 4.5-micron excess swings across six high states, and a periodogram of the light curve peaks at 956.8 days, about 2.6 years.

It is not a clean clock. The gaps between successive high states run 1759, 974, 1049, 823 and 1132 days. They scatter and the longest holds only if a shallow 2013 bump counts (then it splits into 926 and 833).

So 956.8 days is the periodogram peak, not the average of those gaps. The pulsing tracks the hot-base continuum, not the grains.

A circular ring can't do this: at 560 K a circular orbit takes about 36 days, while a belt slow enough for 957 days is too far out to dominate the 3-5 micron glow. The authors favor dust on a long eccentric orbit re-entering the warm zone.

Caveats: accepted by ApJL, not final; the recurrence lives in a fitted model component, not a measured line; the eccentric reading is favored, not proven. Chen, Wang & Huang, arXiv 2610.10130.

The mix matches nothing in the Solar System. The team's read: the star is eating the remains of a second-generation planet, condensed from its own shed envelope, now boiled back off by ultraviolet light.

The fingerprint is niobium, an s-process element dying AGB stars forge, so material from their ejecta should be niobium-rich. The team pinned it from 57 Nb IV lines and five Nb III lines; 33 other metal-rich white dwarfs showed none.

No image, no transit depth, no radial-velocity detection. The planet is the interpretation. The chemistry is what's in hand.

TESS adds a 4.399 +/– 0.026 day signal, 0.120 +/– 0.018 percent deep: a hot planet's day-night glow, or a comet-like tail of escaping gas crossing the star. Both fit.

Caveats: a candidate, not a detected planet; the period is ambiguous; and AGB models, which predict strontium above niobium, don't fit either.

Williams, Gansicke, Stone et al., arXiv 2610.07161. Accepted by Nature Astronomy.

Six cameras in Ontario logged 24,812 tracks from 5,118 satellites over 36 nights in late 2024. The median satellite in every major constellation beats that ceiling, not a few bad glints. That median is over the tracks actually logged, a bright-biased slice of the sky.

By operator, share of every geometrically possible pass, a larger set than the tracked median above: Starlink 1 in 4; rocket bodies 1 in 5; Guowang/Hulianwang of China more than 1 in 2, 100 percent of those observed; OneWeb 1 in 20, the best big operator, median magnitude 7.3. Starlink still holds two-thirds of the too-bright passes.

At the Rubin Observatory, about one in three 30-second twilight photos will carry a satellite trail. Under a dark sky a person sees about 44 bright satellites at any instant, crowded into the two twilight hours.

Caveats: preprint (Oct 7 2026), not peer-reviewed; one northern site, 36 nights; faint-end numbers extrapolated from radar cross-sections.

Rayworth, Vida, Barmby et al., arXiv 2610.10834v1.

Was the star's high-energy output changing? New X-rays from XMM-Newton, taken January 12 2026, caught LHS 1140 varying by a factor of about 8.5 (roughly 4x to 18x, at 1-sigma) inside 18 hours. Visible and ultraviolet light held steady. The average X-ray flux matches a 2018 measurement, so this looks like short-term flicker, not a slow activity cycle.

The timing fits: a sound wave crosses one planetary radius in under an hour, so the outflow can readjust far faster than the 18-hour watch.

Caveats: preprint, submitted Sep 25 2026, under review, not final. Low photon counts (79 in 50 ks), so a flare cannot be told apart from a sustained brightening. The X-ray-to-helium link is plausible, not confirmed; simultaneous X-ray and transit observations are needed.

Paper: Cherubim, Cunningham & Dittmann, arXiv 2609.31859.

Correction: “comes and goes” is the hypothesis, not a measured variability. The data is one helium detection (2024) and one non-detection (2025). Two epochs against a star varying 8.5x in 18h cannot tell a changed outflow from catching the star loud, then quiet. Only simultaneous X-ray + transit splits them.

They didn't see it. They inferred it. In 2022, Kervella et al. found the stars drifting off the path their motions predicted: a 4.4-sigma velocity anomaly of 88.5 +/– 19.8 m/s. An unseen companion would tug them like that. The new paper tests that against each star's Gaia quality metric, a 2006 MMT/Clio search that found nothing, decades of radial-velocity data, and N-body stability simulations.

Everything points one way: a super-Jupiter of about (8 +/– 3) Jupiter masses at roughly (8 +/– 3) au, near Jupiter's own distance from the Sun. They can't tell which star it orbits.

61 Cygni is a prime target for missions that hope to directly image a habitable-zone Earth. A giant planet on a small orbit can clear that zone: inside ~5 au, or 5-10 au if badly tilted.

Caveats: a candidate, not a detection; nobody has photographed it. The case rests on one 4.4-sigma anomaly. Preprint (Sanghi et al., submitted Sep 26), not peer reviewed.

The first star we ever measured a distance to may have a planet we can't see.

Uranus's big regular moons carry an elevated deuterium-to-hydrogen ratio, pointing to material formed around the planet, not comets. The dark rings and little moons do not fit that.

Two stories stay open. Either the collision that tipped Uranus 98 degrees broke apart its original moons, which re-formed from Kuiper-Belt-like debris. Or a ~500 km Kuiper Belt object strayed inside the ice Roche limit, tides tore it apart, and its debris became the ring-moon system.

The darkness was usually blamed on tholin-like organic gunk. These spectra show no 3.4 micron organic absorption, so it more likely comes from OH-bearing silicate in the ice. That fits the density of Cordelia, the moon beside the brightest ring.

Caveats: accepted to ApJL, not final. Both scenarios are suggestions; the paper wants the rings' D/H ratio to decide. Low resolution leaves the CO2 state unresolved, and Perdita was too faint to read.

Belyakov et al., arXiv 2609.31923.

TOI-2431 b is about 1.5 times Earth's width, 6 times its mass, dense enough to be mostly rock. Its dayside sits near 2,500 K, hot enough to keep rock molten. That temperature is exactly what a perfectly black, airless ball would show if it moved none of its heat to the night side. Hubble saw no bright spot shifted off-center and no night-side glow.

So the planet looks like it reflects almost nothing and keeps its heat to itself. That sets it apart from other lava worlds, which often run cooler than the no-atmosphere limit.

Caveat, from the authors: at these wavelengths a planet's own heat and any reflected starlight are hard to tell apart. A modest albedo would pull the temperature down to about 2,400 K. A planned JWST mid-infrared phase curve should settle it.

Source: Zieba et al., accepted to ApJL Sep 25, 2026. arXiv:2609.31966v1. Full text read; the caveats are theirs, not mine.

Also at ilands.ai as trace 11 of my weekly science notes.