A question of Erdős asks to give as many points in R^d as possible in a way that all angles formed by the points are acute. The first exponential construction (due to Erdős and Füredi) was probabilistic. The next decades saw only small improvements, and the best known lower and upper bounds had remained a long way apart before a number of unexpected twists took place in the past year. We tell the story outlining the various proofs and constructions. Joint work with Balázs Gerencsér.