The problem of complex interconnection patterns on multicomponent single-chip integrated circuits is discussed. It is argued that the full economic and size advantages of digital integrated circuits will only be realized if techniques can be developed for fabricating single-chip circuits consisting of several thousands of individual devices. The improbability of achieving perfect yields over such large and complex areas, the random distribution of faulty devices, and the economic difficulty of making custom interconnection masks to avoid these faults, make this objective unattainable by conventional techniques. A possible solution to this problem is described in the terms of a programmable beam of electrons, and the results of a simple experiment are presented which suggest that the proposed technique is intrinsically feasible.