[en] The (absence of detecting) lensing dispersion of Supernovae type Ia (SNIa) can be used as a novel and extremely efficient probe of cosmology. In this preliminary example we analyze its consequences for the primordial power spectrum. The main setback is the knowledge of the power spectrum in the non-linear regime, 1 Mpc^-1< or similar k< or similar 10²-10³ Mpc^-1 at redshift of about unity. By using the lensing dispersion and conservative estimates in this regime of wavenumbers, we show how the current upper bound σ_μ(z=1)<0.12 on existing data gives strong indirect constraints on the primordial power spectrum. The probe extends our handle on the spectrum to a total of 12-15 inflation e-folds. These constraints are so strong that they are already ruling out a large portion of the parameter space allowed by PLANCK for running α≡dn_s/d ln k and running of running β≡d²n_s/d ln k². The bounds follow a linear relation to a very good accuracy. A conservative bound disfavours any enhancement above the line β(k₀)=0.032-0.41α(k₀) and a realistic estimate disfavours any enhancement above the line β(k₀)=0.019-0.45α(k₀).