ABSTRACT We explore the past evolution of Saturn's moons using direct numerical integrations. We find that the past Tethys-Dione 3:2 orbital resonance predicted in standard models likely did not occur, implying that the system is less evolved than previously thought. On the other hand, the orbital inclinations of Tethys, Dione, and Rhea suggest that the system did cross the Dione-Rhea 5:3 resonance, which is closely followed by a Tethys-Dione secular resonance. A clear implication is that either the moons are significantly younger than the planet or their tidal evolution must be extremely slow (Q > 80,000). As an extremely slow evolving system is incompatible with intense tidal heating of Enceladus, we conclude that the moons interior to Titan are not primordial, and we present a plausible scenario for the system's recent formation. We propose that the midsized moons re-accreted from a disk about 100 Myr ago, during which time Titan acquired its significant orbital eccentricity. We speculate that this disk has formed through orbital instability and massive collisions involving the previous generation of Saturn's midsized moons. We identify the solar evection resonance perturbing a pair of midsized moons as the most likely trigger of such an instability. This scenario implies that most craters on the moons interior to Titan must have been formed by planetocentric impactors.