We construct a two-zone model to describe H and He burning on the surface of an accreting neutron star and use it to study the triggering of type I X-ray bursts. Although highly simplified, the model reproduces all of the bursting regimes seen in the more complete global linear stability analysis of Narayan & Heyl (2003), including the regime of delayed mixed bursts. The results are also consistent with observations of type I X-ray bursts. At accretion rates Mdot < 0.1 Mdot_Edd, thermonuclear He burning via the well-known thin-shell thermal instability triggers bursts. As Mdot increases, however, the trigger mechanism evolves from the fast thermal instability to a slowly growing overstability involving both H and He burning. The competition between nuclear heating via the beta-limited CNO cycle and the triple-alpha process on the one hand, and radiative cooling via photon diffusion and emission on the other hand, drives oscillations with a period approximately equal to the H-burning timescale. If these oscillations grow, the gradually rising temperature at the base of the helium layer eventually provokes a thin-shell thermal instability and hence a delayed mixed burst. For Mdot > 0.25 Mdot_Edd, there is no instability or overstability, and there are no bursts. Nearly all other theoretical models predict that bursts should occur for all Mdot < Mdot_Edd, in conflict with both our results and observations. We suggest that this discrepancy arises from the assumed strength of the hot CNO cycle breakout reaction 15O(alpha,gamma)19Ne in these other models. That observations agree much better with the results of Narayan & Heyl and our two-zone model, both of which neglect breakout reactions, may imply that the true 15O(alpha,gamma)19Ne cross section is much smaller than assumed in previous investigations.