GTE (Generative Triple Evolution) coordinates derived from nucleon-triple compositions provide competitive predictive power for nuclear binding energies. In a controlled ablation using exactly 50 features, GTE composition features achieve a 10-fold cross-validated MAE of 3.17~MeV versus 4.24~MeV for an equal-size random polynomial baseline—a ~25\% improvement demonstrating that GTE coordinates carry non-trivial nuclear information beyond arbitrary polynomial combinations of (A,Z). A 6-term analytical law achieves 0.032~MeV/A MAE for binding energy. A 10-term stability classifier using NUBASE2020 labels achieves 94.5\% accuracy in 5-fold cross-validation (majority-class baseline: 75.0\%). Smooth analytical stability laws cannot resolve discrete anomalies such as Tc and Pm; this limitation is disclosed in . Two GTE-theoretic predictions follow from the same arithmetic. First, the nuclear pairing constant Δ_ pair = a_ seed^g/2 = 5^3/2 = 11.18MeV, matching modern SEMF fits within 0.003\% with no free parameters; the underlying GTE b-seed parity is certified in Lean~4 ( .GTE.NuclearPairing, zero sorry). Second, all seven nuclear magic numbers \2,8,20,28,50,82,126\ are derived from the GTE cascade via pion-exchange physics (κ_ emp/κ_ min(N=50) = 1.149 ≈ IPT = 1.1309, 1.6\%, bridge claim). Scope: Stability predictions for Z > 118 are extrapolative ( ).