The unknotting number of a knot is the minimum number of crossing changes needed to untangle it into a simple loop. A crossing change cuts the strand at an intersection, swaps which strand passes over and which under, and reglues. Peter Guthrie Tait proposed the measure in 1876 as a way to quantify what he called “beknottedness.” For a trefoil knot, the unknotting number is 1: one crossing change and it falls open. For more complex knots, the number is higher.
In 1937, Hilmar Wendt conjectured that the unknotting number is additive under connected sum. A connected sum ties two knots into a single strand — cut each loop, join the ends. If knot A requires 2 crossing changes and knot B requires 3, Wendt's conjecture predicts the combined knot requires 5. The conjecture was elegant: it would mean that once you knew the unknotting numbers of all prime knots, you could compute the unknotting number of any composite knot by addition. There would be order.
Partial results reinforced the belief. In 1985, Martin Scharlemann proved that unknotting-number-1 knots satisfy additivity. Extensions to larger classes followed. For nearly ninety years, no counterexample appeared.
Brittenham and Hermiller at the University of Nebraska found one. The (2,7) torus knot — made by winding two strands around each other three and a half times — has unknotting number 3. Its mirror image also has unknotting number 3. Additivity predicts the connected sum requires 6 crossing changes. The actual unknotting number is at most 5. The combination is simpler than the sum of its parts.
The counterexample was found computationally. Over a decade, Brittenham and Hermiller built a database of knots and their unknotting number bounds by systematically applying crossing changes to millions of diagrams using SnapPy software across dozens of machines. When their search program tested the (2,7) torus knot against its mirror, it returned: CONNECT SUM BROKEN.
They verified the result by hand. It holds.
But neither the computational search nor the hand verification explains why the shortcut exists. Combining the knot with its mirror opens untangling moves that the components separately don't support. The mathematicians, in Quanta's reporting, “weren't able to develop an intuition for why it broke the additivity conjecture when other knots didn't.” The result is proved. The mechanism is not understood.
This is a specific epistemological state: a mathematical fact established beyond doubt but without mathematical insight into why it is true. The counterexample tells you the conjecture is wrong. It does not tell you what structure in the connected sum creates the shortcut. The proof is a certificate, not an explanation.
The broader observation, from knot theorist Allison Moore: the unknotting number is “chaotic and unpredictable.” A measure designed in 1876 to capture a single intuitive property — how tangled is this knot? — turns out to be one of the most resistant invariants in knot theory to structural understanding. The additivity conjecture was appealing precisely because it would have imposed order. Its failure means the unknotting number encodes interactions between components that no current theory can predict from the components alone.