Photoreal Solar System

I have three children, and they got to astronomy before I did. It started with the usual questions — why the Moon follows the car, why Saturn has rings, whether you could stand on Jupiter — and I ran out of good answers fast. Every app I could find to show them either drew the planets in a tidy row at the wrong sizes, or put them at the right sizes and left the screen empty, because that is what space actually looks like. So we started building the thing we wanted to look at.

I built it with AI — Claude wrote a great deal of this code, working from what I asked for and what we measured. I want that said plainly and up front rather than discovered later. What it did not do is invent the astronomy: every position here comes from the same ephemerides an observatory uses, checked against JPL Horizons, and every surface is real NASA, ESA and USGS imagery. When the numbers disagreed with the render, the render was wrong and got fixed.

My children have been the reason this took the shape it did. They asked to land on the Moon, so it streams real imagery down to a third of a metre per pixel over the Apollo 12 site. They wanted to see an eclipse, so the Sun became a disc rather than a point and the shadows are computed properly, with a real umbra and penumbra. They kept asking to go further in, and every time they did, something had to be rebuilt.

It is free and there is no account. I do count visits — how many people arrive and roughly where from — because otherwise I have no idea whether anyone is here at all. No cookie, no profile that outlives the day, nothing sold, and nothing that follows you off this page. The one thing it costs me is bandwidth — all that imagery is served on demand and paid for out of pocket. If it has given you or your own kids a good half hour, there is a Support tab in the bar. Entirely optional, and the whole thing works the same either way.

Textures are streamed NASA/ESA/USGS tiles. The Moon goes to 0.325 m/px over the Apollo 12 landing site on LRO NAC, with the terrain actually displaced by LOLA elevation, not bump-mapped. 875,291 real stars, and the Milky Way behind them is a photometrically anchored plate. That last one had a 90° pole-axis error plus an east-west mirror for months and rendered 10-27x too faint in Sagittarius. I only caught it because I measured the rendered frame against the drawn star field instead of trusting my eyes.

Self-imposed rule: nothing gets drawn that isn't measured, unless the screen admits it. Bodies use their own photometric function, not a cosine. Saturn's rings are Voyager PPS optical depth at 5 km.

Black holes: Sagittarius A, M87, Cygnus X-1, with published masses and distances and their real error bars, asymmetric where the paper published them asymmetric. Two of the three say "not measured" under spin, because nobody has.