TWO-MINUTE EXPLAINER
The idea, as a cartoon.
What Reality Compute is testing, why most results so far are nulls, how every contribution is scored, and how you can help.
Animated in code and narrated by Apple’s “Zoe (Premium)” synthetic voice, with a quiet music bed. Captions are on by default and can be turned off in the player. The opening image is the Webb First Deep Field (NASA, ESA, CSA and STScI, public domain).
Transcript
- Thousands of galaxies, in a speck of sky the size of a grain of sand. Could all of it be running on something like a computer?
On screen: The camera slowly pushes into the Webb First Deep Field, thousands of galaxies on black; the words “Is all of it computed?” rise in gold. - Reality Compute is an open investigation into that question.
On screen: A small glowing universe sits on the screen of a friendly desktop computer, with a question mark bobbing beside it. - It is a fun idea. But it only becomes science if it makes a prediction we could actually check.
On screen: A thought bubble shows the simulation idea; a checklist card appears and ticks “Makes a prediction?”. - A perfect simulation could never be spotted, because every experiment would come out the same. So we look for the kind that could leave a trace: a computer with a limited budget, that has to cut corners when the work gets hard.
On screen: Two computers side by side: a calm “Perfect copy” showing two identical worlds, and a sweating “Limited budget” machine whose gauge swings towards red while scissors cut corners. - Here is the test. Take two quantum circuits that look like twins to ordinary noise, but one is far harder to simulate. If a hidden computer were straining, the hard twin would come out slightly worse.
On screen: Two quantum-circuit diagrams labelled “Easy to simulate” and “Hard to simulate” sit above a balance; a magnifier finds a tiny error blip on the hard one. - So far, every test has come out null. No sign of strain. That is not a failure. Each result narrows down what kind of computer could be running things, and each one is published with its limits.
On screen: A chart of test results sits inside a shaded “no effect” band; a “No strain found” stamp lands, and a funnel narrows the possibilities. - Nothing here shows that reality is simulated. The research is about what would count as evidence, and what would rule it out.
On screen: A magnifying glass passes over a card titled “Evidence?” with two columns: what would count, and what would rule it out. - Every contribution gets two scores: how much it changes the theory, and how much closer it brings an answer, whichever way that answer turns out. Watch the total climb on the homepage.
On screen: Two cards: a colour bar from red-black (0, no relevance) to bright green (100, breakthrough) with a marker sliding along it, and a line that climbs in small steps toward 100, “progress to an answer”. - You can help. Share a paper or a video. Challenge a claim. Propose a test. Or point your own AI at an open question. Every contribution is credited, checked, and can change the theory.
On screen: Four icons appear — share a source, challenge a claim, propose a test, bring your AI — linked by a loop: contribution, checked, theory updated. - Reality Compute. An open investigation, at reality-compute.fyi.
On screen: The end card shows three photographs — the Webb deep field, Meteor Crater in Arizona and a fractal romanesco — above reality-compute.fyi, “Share a link, a video or an idea” and “Nothing here proves reality is simulated.”
What the video does not claim
It does not claim that reality is simulated, and it is not evidence either way. It simplifies: the real tests, their error rates and their limits are in the executive summary and the version history.