If our universe is a simulation running on a grid, its screen resolution should leave fingerprints. That is not a stoner thought experiment. In 2012 three working nuclear physicists — Silas Beane, then at the University of Bonn, with Zohreh Davoudi and Martin Savage at the University of Washington — published a paper titled 'Constraints on the Universe as a Numerical Simulation' that turned the simulation hypothesis into something almost nobody had managed before: a falsifiable prediction with a number attached.

Their reasoning starts from what physicists actually do for a living. When you want to simulate the strong nuclear force — quantum chromodynamics, the math that binds quarks into protons — you cannot compute it in smooth, continuous space. The equations are too violent. So you do it on a lattice: you replace continuous spacetime with a finite cubic grid of points, lattice QCD, and let the physics evolve in discrete steps. Beane and his coauthors asked the obvious inverted question. Suppose a far more advanced civilization were simulating an entire universe — ours — using the same trick we already use, a cubic spacetime lattice. What would living inside that grid feel like, and could the inhabitants ever detect the pixels?

The key insight is that a grid is not perfectly smooth, and a grid has a smallest possible scale: the spacing between adjacent points. Nothing in the simulation can be finer than one pixel. That hard floor imposes a maximum possible energy on any particle — and here is the part that makes the paper testable. Real space has no preferred directions; you can rotate freely and physics looks the same. A cubic lattice does not. It has edges and diagonals; moving along an axis is subtly different from moving corner-to-corner. So at the very highest energies, where particles would 'feel' the granularity, this hidden cubic scaffolding should betray itself as a tiny breaking of rotational symmetry — a faint preference for certain directions baked into the structure of space itself.

Then they pointed to where the fingerprint would already be sitting in the data. The highest-energy particles we know of are ultra-high-energy cosmic rays, and their spectrum has a real, observed cliff called the GZK cutoff — a sharp drop-off in cosmic rays above roughly 10^19 to 10^20 electron-volts, caused by those particles colliding with cosmic microwave background photons. Beane's team noted that this same cutoff energy gives the most stringent bound on any underlying lattice spacing: if the universe is a grid, its inverse spacing must be at least about 10^11 GeV. And crucially, if the GZK edge is set by a lattice rather than by ordinary physics, then the arrival directions of the most energetic cosmic rays should not be perfectly uniform across the sky. They should cluster, ever so slightly, along the axes of the cosmic grid — a directional bias pointing straight at the rendering engine.

The honest caveats are large, and the authors stated them plainly. Their result assumes the simulators use the same kind of lattice we use, with the same kind of physics on top. A genuinely alien architecture — a different grid geometry, a randomized mesh, an adaptive resolution that refines itself only where someone is looking — would erase the cubic signature entirely and leave their test blind. The bound is also a floor, not a detection: it says only that if there is a lattice, the pixels are smaller than a staggeringly tiny length, far beyond current reach. Nobody has measured rotational symmetry breaking in cosmic rays. The paper proves nothing about whether we are simulated.

What it does, and why it endures, is convert metaphysics into methodology. Most simulation arguments are unkillable because they predict nothing — any evidence can be waved away as 'that's just how the program is written.' Beane, Davoudi and Savage refused that escape hatch. They said: if reality is a particular kind of computation, the computation has a resolution, the resolution has consequences, and the consequences land in a dataset we are already collecting. Either the highest cosmic rays show a cubic anisotropy or they don't.

The unresolved part is the most unsettling. The experiment is real and the prediction is sharp, but the sensitivity needed to confirm or rule out a lattice at the GZK scale is currently out of reach, and a sufficiently clever simulator could simply choose an architecture with no straight edges to find. So the test cuts both ways. A null result never fully exonerates the hypothesis — it only rules out the lazy version of the grid. Which leaves the question the physicists handed us and walked away from: we now know how to look for the pixels. We just can't yet tell whether we're failing to find them, or whether something made sure we never would.