The computational hardness conjecture for constant discrepancy

Let $A$ be drawn from the $(m,n,1/2)$-Bernoulli ensemble, where $C$ is a universal constant and $n\geq C m\log m$. Computational hardness conjecture. With high probability, no efficient algorithm finds a solution of constant discrepancy for $A$.The conjecture is motivated by the planted model, where clusters of low-discrepancy solutions are isolated from one another. It asserts an algorithmic barrier for finding constant-discrepancy solutions in the stated random regime.