The three "soften-with-attribution" items already contain "reportedly" in the article, and the reviewer notes explicitly conclude "No issue"/"No problem"/"This checks out" for them. The fourth item ("remove") refers to a sentence from the evidence pack (E9), not the article, and is marked "Disregard." Therefore no changes apply to the article.
Article — "ESO warns 1.7 million proposed satellites could devastate ground astronomy as Amazon accelerates": Article — "ESO warns 1.7 million proposed satellites could devastate ground astronomy as Amazon accelerates": Two men who both love space cannot agree on whether it is crowded.
One of them, Elon Musk, said last month, "Space is really big, so it's not like space is going to get crowded." He said it in a video announcing plans to put up to a million satellites in orbit as space-based data centers.
The other, Olivier Hainaut, has spent more than thirty years at the European Southern Observatory watching the sky fill up. His answer arrived as a peer-reviewed number: no more than 100,000 satellites, all of them too faint to see with the naked eye.
That is the collision this week. The largest camera ever built sits at the Vera C. Rubin Observatory in Chile. And a single company reportedly wants to fly ten times more satellites over it than the whole planet can bear.
This week Hainaut's study, accepted for publication in Astronomy & Astrophysics, turns that argument into a hard engineering threshold, and it lands just as Amazon accelerates its own deployment against a regulatory clock.
Start with the number nobody in the industry wanted computed. There are about 14,000 satellites in orbit today, most of them SpaceX Starlinks. The proposals on file now exceed 1.7 million.
Hainaut's ceiling is 100,000. That is more than seven times today's population, so this is not a demand to stop. It is a demand to stop at a level an order of magnitude below what one company alone has announced.
Hainaut's ceiling is 100,000—more than seven times today's population, yet an order of magnitude below what one company alone has announced.
Here is the part that reframes the whole fight. The problem was never just the streaks.
Everyone has seen the images: a satellite crosses the frame, the Sun lights it up, and it leaves a bright trail that "zaps whatever is behind it," as Hainaut puts it. Those trails you can, in principle, mask and subtract. Astronomers have been managing.
The study's real contribution is measuring the second effect—the one you cannot subtract. Satellites too dim to see individually still throw off a veil of diffuse light. Brighter ones scatter light through the atmosphere in every direction. Both raise the background brightness of the entire sky.
And sky brightness is not cosmetic. For any faint target, the exposure time needed to reach a given signal-to-noise ratio scales linearly with how bright the background is. Double the sky glow and you double the time to see the same distant galaxy, per the study's linear-scaling relation.
Professional observatories have already retreated to the darkest sites on Earth precisely to keep light pollution below one percent of the natural sky. A ten percent rise is, per the literature, incompatible with state-of-the-art work.
This is why one company's promise deserves close reading. Reflect Orbital, the California startup that wants to fly giant orbital mirrors to beam sunlight down at night, told AFP it will "systematically avoid redirecting light near observatories."
That pledge answers the wrong question. Aiming is about where the beam points. Scattered background light does not care where the beam points—it pollutes the whole sky regardless of aim.
The physics here is brutal. Hainaut ran his simulation assuming no Reflect Orbital satellite ever points its beam at or near an observatory. Even so, the trail from a single mirror-satellite could ruin a Rubin exposure.
With the full 50,000-satellite fleet, every image from a camera like Rubin's would be lost whenever the satellites are sunlit. And each of those satellites, even when its beam is aimed elsewhere, would reportedly shine as bright as Venus. Aim it at you and it appears four times brighter than the full Moon.
The whole night sky, the study reportedly finds, would end up three to four times brighter with that fleet up. Hainaut's plain-language version: whether you stood in France, the Sahara, or Chile, the sky "would no longer be clear, resembling instead the sky seen in the suburbs of a city." From a light-polluted city like Munich, the satellites would be the only "stars" visible at all.
So the "we'll aim away" guarantee is not dishonest. It is simply impossible to keep in the way that matters.
Now the megaconstellation. For SpaceX's million-satellite data-center swarm, Hainaut found that two hours into the night, dozens of trails would cross each VLT image—field-of-view losses up to 28 percent. That assumes the satellites stay faint enough to be naked-eye invisible.
Make them just a little brighter and the saturating cameras break. Rubin's could have most of its images rendered unusable for several hours a night.
Against all of this sits Musk's "space is really big." As an intuition about volume, it is not wrong. As a claim about the sky, it misses that we do not observe volume—we observe lines of sight, and a satellite anywhere along that line, however sparse the traffic, sits between the telescope and the universe.
What makes Hainaut credible is that he refuses to oversell his own number. "This is not a hard number, like 99,999 is good and 100,001 is bad," he said. "Clearly I'd prefer 50,000. But 100,000 causes losses at about the level of other technical losses, such as equipment failure."
That is the tell of an honest threshold. It is not a red line drawn for drama. It is the point where satellite damage disappears into the noise of downtime observatories already tolerate—weather, broken instruments, ordinary bad luck.
Which makes 100,000 a livable compromise, technically. Sixty thousand satellites that all stay fainter than magnitude 7 contribute, per the study, only about a ten-thousandth of the natural dark sky. The industry could hit that if it were forced to.
But nothing in the current structure forces it. This study exists specifically to arm ESO's response, filed with the Royal Astronomical Society and the International Astronomical Union, to the FCC applications from SpaceX and Reflect Orbital.
The FCC is a national broadband regulator. Its instincts reportedly run toward buildout, not restraint. And its clock is running the wrong way: Amazon reportedly faces a July 2026 deadline to have half its roughly 3,200-satellite Leo constellation operational, with only about 370 launched so far.
That deadline pushes Amazon and its rivals to launch faster, not slower. The body ESO's institutional affairs officer Betty Kioko calls the holder of an "existential" decision has every institutional incentive pointed at the opposite outcome.
Meanwhile the sky is already crowded in a second sense: reportedly 13,000 tons of debris up there, with the standing risk that collisions cascade into the runaway chain reaction called Kessler syndrome.
So this is a governance test disguised as an engineering paper. The number is reachable. The physics is settled. What is missing is any mechanism—at the FCC, in Musk's math, or in orbit—built to enforce a ceiling that everyone, once they read it, can see is generous.