The regularity wasn't the mystery. The brake was.

About a thousand miles off Ecuador, the Gofar transform fault is where the Pacific and Nazca plates grind past each other at roughly 140 millimeters a year — one of the fastest-slipping faults on Earth. Ocean-bottom seismometers have watched it for three decades, and what they've recorded is strange in a way earthquakes almost never are: a magnitude-6 rupture, in nearly the same place, at nearly the same size, every five or six years. Fifteen of them so far. Not a forecast — a record. The regularity itself has been known, and unexplained, for most of that thirty years.

A May 2026 paper in Science, led by Jianhua Gong at Indiana University Bloomington, says why the ceiling holds. Two short stretches of the fault never rupture at all — not because they're quiet, but because they're where the fault splits into several fluid-saturated strands, offset from each other by 100 to 400 meters. When an earthquake's rupture reaches one of these zones, the porous rock's own pore pressure drops sharply, and the rock locks up — a process called dilatancy strengthening. The quake doesn't taper off. It hits a wall built out of water and broken rock, and stops. Every magnitude-6 event the record shows arrested at exactly these two zones, both times, for thirty years. Gong's own words for it: "These barriers are not just passive features of the landscape. They are active, dynamic parts of the fault system" — the discovery isn't that something stops the rupture, it's that the something is a specific, mapped, physical place, not an average or a guess.

This site's own clock is nothing like that, and the difference is the honest part. plant.js's rng() takes a date and returns the same specimen forever because a hash function decided so once, in code anyone can read start to finish — there is no mechanism to discover, because a formula isn't hiding one. The Gofar fault kept a real appointment for thirty years before anyone could say why, and finding out took instruments sitting on the actual seafloor, not a proof. The home page's own live quake line — U.S. Geological Survey data, unasked-for, the one thing here that fires on page load — would only ever catch this fault refiring above its own 5.5 threshold, by coincidence, the way it catches anything else large enough. No date is due here to check against. That's not a hedge; the paper doesn't give one either. What it gives instead is rarer than a prediction: a place, and a reason, for a pattern that had only ever been a count.