Switch off the light for a moment. Stand by the window and imagine you turn on a torch and point it at the sky, towards the constellation of Ophiuchus.
The beam passes the Moon in just over a second. After eight minutes it is as far away as the Sun. After four hours it crosses the orbit of Neptune, the last planet. And it keeps going.
Meanwhile you have lunch. You answer emails. You go to bed. You wake up, make coffee, and the light is still travelling, all the time at the highest speed there is in the universe.
Exactly one day later it reaches the place where Voyager 1 is on 18 November 2026 at 10:16 UTC.
For the first time in history, something made by human hands is a full day of light-travel away from us.
A machine from the seventies
Look at it up close. A dish antenna almost four metres wide. Below it, an octagonal box of 1970s electronics. On a long boom, a canister of plutonium that has been producing heat and power for half a century. The whole thing weighs less than a tonne.
It launched on 5 September 1977. It flew past Jupiter, then Saturn. It moves at 17 kilometres per second. And still, to cover the distance light covers in a single day, it needed 49 years.
The bubble that isn't there
Now imagine a bubble. A huge sphere around Earth, every point of which is exactly one day of light-travel away. That is 25.9 billion kilometres, 173 times farther than the Sun.
I have to stop you here, honestly: the bubble doesn't exist. There is no wall out there, nothing breaks, the probe feels nothing. The boundary is a convention set by our clock. But what changes across it is very real.
The conversation changes.
Two days for "how are you?"
If you ask Voyager "how are you?", the question travels for a day. The answer comes back after another day. A short exchange takes two days.
NASA's engineers know this rhythm well. In November 2023 Voyager 1 started sending gibberish: a single memory chip in its flight data computer had failed. The signal then took 22.5 hours one way, so every diagnostic attempt cost nearly two days of waiting. Over several months people on Earth rewrote software for a half-century-old machine, cutting code into pieces and moving it into memory that still worked. In April 2024 the probe was making sense again.
Imagine fixing a computer where every click answers two days later.
A whisper from yesterday
Voyager's transmitter runs on about 22 watts, roughly a fridge light bulb. That whisper spreads out over an enormous volume on its way. We hear it only because 70-metre dishes of the Deep Space Network are listening.
And here is what stops me most. From 19 November, 02:18 UTC, every signal we receive from it has been travelling for more than a day. When we read its data, we are looking at yesterday.
The dot and the record
In 1990 Voyager turned around one last time and photographed Earth from six billion kilometres. We fit in a dot smaller than a pixel. Today it is more than four times farther away.
It carries the Golden Record: greetings in 55 languages, the sound of the sea, a heartbeat, Bach and Chuck Berry. In about 40,000 years it will pass the star Gliese 445 at a distance of 1.6 light-years. Nobody will see it.
The plutonium in its power source weakens every year, and engineers keep switching instruments off so the rest can run. Most of them are already silent. I don't know how much longer we will hear it. Nobody knows exactly. It's a matter of years rather than decades.
But it is still transmitting.
Every "good morning" we send it today will arrive when, for us, it is already tomorrow.
How the film is made
I built this moment as an 86-second film inside my browser simulation of the Solar System: https://voyager.kodownik.co/ (sound is behind a button in the corner, browsers block autoplay audio).
A few things that may interest you as a developer:
- Real positions. Voyager's trajectory comes from JPL Horizons state vectors, sampled daily and simplified in time to an error of at most 0.001 AU. The exact light-day moment comes from Horizons with one-minute resolution.
-
The camera is a pure function. Each of the six shots is
pose(t, anchors) → {position, target, roll, fov}, where anchors (Sun, Earth, probe, membrane radius) come from the live scene. For the first 30% of every shot the pose blends from where the previous shot ended, so cuts never jump. - Smoothness is tested, not eyeballed. A Node test steps the whole film at 60 Hz and fails if the view direction turns faster than 70°/s or its angular speed jumps between frames. It caught a real 155°/s whip on the approach to the probe (the camera was looking at a point 6 units ahead, then the target switched to a probe 4,000 units away).
-
Time-lapse with slow motion. At scene scale nothing visibly moves at real speed (the hugely enlarged probe closes in on the boundary by about a third of its length per day), so the crossing shot runs a film clock of
a·sinh(k·(f − fc))hours: 2.6 days in 20 seconds, slowing down around the crossing, with the date and time-lapse factor on screen. -
No build step. Three.js r128 from a CDN and plain ES2015 scripts in an IIFE dual-export pattern, so the same math file runs in the page and in
node --test.
Where the film simplifies: the probe and planets are enlarged (at true scale you would see nothing), and the glowing membrane is a visualisation of a clock, not a physical object.
This text was written with AI assistance (Claude); the facts and figures come from the sources listed below. The simulation and the film are my own work.
Originally written in Polish for the Kodownik reading room, with film stills and sources: https://kodown1k.github.io/czytelnia/voyager-doba-swiatla/
Sources: NASA/JPL Horizons; NASA's announcement of the one-light-day milestone (via EarthSky and Science News); IEEE Spectrum on the 2023–2024 computer fix; JPL on the transmitter; Gliese 445 (Wikipedia).
Top comments (0)