The high pT > 3 GeV azimuthal asymmetry, v2(pT), in non-central nuclear collisions at RHIC is shown to be a sensitive measure of the initial parton density distribution of the produced quark-gluon plasma. A generalization of the Gyulassy-Levai-Vitev (GLV) non-abelian energy loss formalism including Bjorken 1+1D expansion as well as important kinematic constraints is used.PACS numbers: 12.38. Mh; 24.85.+p; Introduction. In order to interpret data on nuclear collisions from recent Relativistic Heavy Ion Collider (RHIC) experiments [ 1,2,3], it is obviously necessary to have knowledge of the initial conditions. Currently, there is an order of magnitude uncertainty in the initial produced gluon density, ρ g (τ 0 ) ∼ 10 − 100/fm 3 , in central Au + Au at √ s = 130 AGeV since widely different models [ 4,5] are consistent [ 6] with PHOBOS data [ 1]. We note that recent PHENIX data [ 3] appear to be inconsistent with at least one class (final state [ 5]) of gluon saturation models. It is essential, however, to check this with other observables as well. High p T observables are ideally suited for this task because they provide a measure [ 4] of the total energy loss, ∆E, of fast partons, resulting from medium induced non-abelian radiation along their path [ 7,8]. For intermediate jet energies (E < 20 GeV), the predicted [ 9, 10] gluon energy loss in a static plasma of density ρ g and thickness, L is approximately ∆E GLV ∼ E(L/6 fm) 2 ρ g /(10/fm 3 ). The approximate linear dependence of ∆E on ρ g is the key that enables high p T observables to convey information about the initial conditions. However, ∆E also depends non-linearly on the geometry, L, of the plasma and therefore differential observables which have well controlled geometric dependences are also highly desirable.