Mark Brittenham and Susan Hermiller ran computations for over a decade, on university supercomputers and old laptops bought at auction, searching for a knot whose unknotting number violates the additivity conjecture. The conjecture, implicit since 1937, states that combining two knots produces a knot whose complexity equals the sum of the parts. It seemed obviously true: if knot A requires three crossing changes to untie and knot B requires three, their combination should require six.
The (2,7) torus knot and its mirror image each have unknotting number 3. The conjecture predicts their connected sum has unknotting number 6. Brittenham and Hermiller showed it's at most 5. One fewer crossing change than expected. The combination is simpler than the sum of its parts.
When Brittenham checked the output files, he found a line reading “CONNECT SUM BROKEN” — a message he and Hermiller had programmed as a joke, never expecting to trigger it. Hermiller's reaction: “We just dropped absolutely everything else. All of life just went away. Eating, sleeping got annoying.” They tied the knot by hand to verify.
The structural observation is about what kind of knowledge a counterexample produces. The noperthedron disproves Rupert's conjecture by exhaustively verifying that a specific shape lacks a specific property. The proof closes a question: the conjecture is false, the counterexample demonstrates it, done. But the unknotting number counterexample opens questions. Allison Moore called it “like waving a flag that says, we don't understand this.” Charles Livingston noted it reveals “much more complexity and unknowns about knot theory than we knew there were a few months ago.” The counterexample didn't just break the conjecture — it exposed that the unknotting number is “chaotic and unpredictable,” a characterization that didn't exist before the counterexample.
Destructive verification — showing something fails — can be more generative than constructive verification. When a conjecture is confirmed, the field contracts: one fewer open question. When a conjecture is disproved by a surprising counterexample, the field expands: the broken assumption reveals structure that the assumption was hiding. The additivity conjecture was a simplifying assumption. Its failure means the unknotting number has richer behavior than “well-behaved and additive.” That richness was invisible while the assumption stood. Breaking the conjecture was the discovery.