Overview
Sometime in late 2024, short videos began appearing on Instagram and TikTok describing a coming gravitational event. The details varied, but the shape was consistent: a program inside NASA had modeled a moment when Earth's gravitational field would briefly fail, the agency had chosen not to warn the public, and the date was fixed. By early 2025 the narrative had acquired a name — Project Anchor — and by 2026 it had acquired a deadline. On August 12, 2026, according to the circulating version, gravity would switch off for seven seconds.
The consequences described in the videos were vivid and specific. People, vehicles, and unsecured objects would rise; buildings would shear; everything would return at once. Casualty figures in the tens of millions were attached to the claim in some retellings, always presented as NASA's own internal projection rather than the poster's guess.
The date was not invented. A total solar eclipse does occur on August 12, 2026. That single verifiable fact is what carried the rest.
The Anatomy of the Claim
The strongest versions of the story worked by layering. First came the eclipse — checkable, calendared, and easy for a skeptical viewer to confirm in seconds. Then came a mechanism: the alignment of Sun, Moon, and Earth would produce a form of gravitational interference, described in some posts as a resonance and in others as an interaction with gravitational waves from distant colliding black holes. Then came the code name, which gave the story an institutional shape. Then came the document.
The document is where the chain ends. Fact-checkers who searched NASA's Technical Reports Server and its public archives found nothing matching any of the circulating descriptions of Project Anchor, and no version of the claim has ever been accompanied by a scan, a file number, or an identifiable author. What circulated instead were screenshots of text — sometimes formatted to resemble a memo, sometimes simply narrated over stock footage of the solar system.
What NASA Said
A NASA spokesperson addressed the claim directly, stating that Earth will not lose gravity on August 12, 2026, and explaining the reason in terms that leave little room for a mechanism. Earth's gravitational force is determined by its mass — the combined mass of core, mantle, crust, ocean, terrestrial water, and atmosphere. For gravity to switch off, that mass would have to go somewhere. Alignments do not remove mass, and eclipses in particular are geometric events: the Moon's shadow crosses the surface, and nothing about the arrangement changes what any of the three bodies weigh.
The gravitational-wave version of the claim runs into a separate problem. Gravitational waves from merging black holes do reach Earth and have been measured since 2015, but the distortions they produce are on the order of a fraction of the width of a proton across a four-kilometer detector. They do not cancel a planet's field; they ripple through it at a scale that required decades of engineering to detect at all.
The Eclipse It Borrowed
The August 12, 2026 eclipse has a magnitude of about 1.04. The path of totality crosses the Arctic, Greenland, Iceland, the Atlantic, and then northern Spain and the extreme northeastern edge of Portugal — the first total eclipse visible from mainland Europe since 1999. That last detail meant the date was already circulating widely in astronomy coverage, travel planning, and eclipse-chasing communities well before the gravity claim attached itself to it.
This is a recurring pattern. Eclipses arrive with a fixed date, global attention, and a built-in sense of something being wrong with the sky, which makes them unusually hospitable to predictions of catastrophe. The 1919 eclipse drew end-of-the-world expectation; the April 2024 North American eclipse drew a cluster of simulation and reset claims. Project Anchor is the same reflex pointed at a European path.
Why the Framing Held
Several features of the claim made it durable. The seven-second duration is short enough to be survivable, which turns the story from an extinction prophecy into something closer to a warning worth sharing. The code name implies compartmentalization, so the absence of evidence reads as confirmation rather than contradiction. And the date sits far enough ahead to be repeatedly re-shared without ever being falsified — a property the claim retained for roughly eighteen months.
That property expires on August 12, 2026. What happens to the theory afterward is the part that has no precedent to draw on, though the usual pattern for dated predictions is not disappearance but revision: the date moves, or the event is said to have occurred and been concealed.