Universal quantum computation requires single-qubit control together with at least one entangling two-qubit gate. CNOT and controlled-Hadamard gates are two famous examples of such a gate, wherein the state of one qubit (control) dictates the transformation applied to the other (target). By using a simple derivation motivated by symmetry, I will show that current device architecture of semiconductor-based quantum dot devices makes it impossible to realize a CNOT or any other control-target gate efficiently via Heisenberg exchange, Coulomb repulsion, or other interaction that is invariant under spin exchange. Guided by this general principle, we propose a novel blueprint of double quantum dot devices, that enables efficient implementation of both the CNOT and the controlled Hadamard. The best part: our numerical simulations predict this novel device blueprint can enable fault-tolerant 1us CNOT gate in isotopically purified silicon.