In the ANTS (Ants Nearby Treasure Search) problem, multiple searchers, starting from a central location, search for a treasure. The searchers cannot communicate and have few bits of initial knowledge, called advice, when they begin the search. In this paper, we initiate the study of ANTS in the geometric plane.Our main result is an algorithm, GoldenFA, that tolerates arbitrarily many crash failures caused by an adaptive adversary, and requires no bits of advice. GoldenFA takes O L + L 2 (t+1) ND log L expected time to find the shape, for a shape of diameter D, at distance L from the central location, with N searchers, t < N of which suffer adversarial crash-failures.We complement our algorithm with a lower bound, showing that it is within logarithmic factors of optimal. Additionally, we empirically test GoldenFA, and a related heuristic, and find that the heuristic is consistently faster than the state-of-the-art. Our algorithms and analysis make critical use of the Golden Ratio.