Sums of matrix absolute values. An Evidence Press briefing. For ordinary numbers, the size of a sum is never more than the sum of the sizes. Matrices also have a natural absolute value, but for matrices this rule can fail: a sum of matrices can be bigger than the sum of their absolute values, even when size is measured by a standard norm. The question is how much bigger, at worst. The answer is a constant, and the task is to find the smallest one that always works. In 2010 Eun-Young Lee asked for this constant for two matrices, using the Schatten p-norms, the standard family of matrix sizes. The case p equals two was solved in 2022. More recently the question was extended to many matrices, and in 2026 a conjectured formula was proved for every p of at least two. Between one and two, several groups showed that the formula is wrong, and a recent survey lists that range as open. This release works in that range. First, when there are at least as many matrices as their size, it finds the exact constant, a maximum of one explicit function, and describes every extreme case. For rank-one matrices, the building blocks of all the known counterexamples, the problem is now solved in every dimension. Second, it gives a new upper bound for any number of matrices. For two matrices at p equals three halves, a sum can exceed the sum of absolute values by at most about six and a half per cent, instead of about twelve per cent before, while some pairs are known to exceed it by three point six per cent or more, so the true answer lies in between. Third, a computer-assisted proof in exact arithmetic shows that two three-by-three matrices beat every rank-one example, at nine exponents between one point one and one point seven. The limits matter. The exact constant for two matrices of any size is still open, the new upper bound is not sharp, and the certified gap is tiny and covers nine exponents only. This is an anonymous, AI-assisted, unrefereed candidate, published on the fifteenth of November twenty twenty-six. Its review was internal and AI-assisted, not peer review. The paper and its evidence are linked from the Evidence Press release page. This briefing uses an AI-generated voice and is not additional scientific evidence.