This paper proposes a unified two-algorithm model of nonlinear dynamics applicable to physical, biological, economic, and cognitive systems. The model postulates the interaction of two competing processes - Algorithm A (a deterministic attractor, a stability vector) and Algorithm B (a stochastic generator, a deviation impulse) as a sufficient mechanism to explain the threshold avalanche transitions observed across a broad class of phenomena: from avalanche breakdown in semiconductors and the neuronal action potential to the collapse of supernovae, from self-organized criticality to evolutionary leaps - that is, across scales from nanometers to light-years. The central contribution of the paper is the introduction of the concept of conditional intent (the IF→THEN operator) as the mechanism by which an observer supplies the controlling system with a directing vector for the avalanche - specifying not merely a goal, but a complete map of the cascade's unfolding. It is shown that this model is consistent with a number of Nobel-winning discoveries in non-equilibrium thermodynamics and the physics of complex systems, and that it also connects substantively with the independently developed program of early-warning signals for critical transitions in ecology and climatology. The two-algorithm model developed here received empirical testing against historical data of the CBOE VIX volatility index for the period 1990-2026. Statistical analysis confirmed the existence of a pronounced asymmetry between endogenous avalanches (preceded by an anomalously prolonged latent stagnation of the system, averaging 82 days) and exogenous shocks (occurring with zero preceding calm). The differences found are statistically significant at the p < 0.05 level (exact Mann-Whitney test, p ≈ 0.029), which supports the plausibility of the proposed two-algorithm model using real macroeconomic processes as an example; the authors explicitly note the limitations of the sample and the associated ceiling on the test's power (Section 8.3), and invite independent replication on larger datasets.