While remarkable experimental developments in the past two decades have managed to improve by nearly six orders of magnitude the coherence time of superconducting qubits, it is commonly accepted that still more gains are needed in order to use them in fault tolerant quantum computation protocols. Therefore, it remains essential to keep developing strategies in order to reduce the influence of environmental noise on such circuits. In this talk, I will give a review of some of the existing proposals to achieve better qubit protection. Using the paradigmatic example of grid states first emphasized by Gottesman, Kitaev and Preskill in 2001, I will outline the guiding principles used in the conception of protected qubits, and present some of the recent circuit designs that have been inspired by them. Then, drawing on analogies between qubit protection and quantum error protection, I will explain how combining qubits in some small arrays can increase significantly their coherence time. A paradigmatic example of such construction is the Kitaev toric code model, which can be implemented by the tool box of Josephson junction arrays, as we have shown already two decades ago in joint works with Lev Ioffe. More recently, the associated quantum surface code has been implemented on a 72 qubit device, illustrating that existing qubits may be at threshold to start seeing improvement due to first steps of quantum error protection.