We develop a sentiment-aware portfolio optimization model and integrate Ant Colony N Optimization (ACO) into the Mamba neural architecture. Our model adapts to nonlinear dynamics, and it is based on a RuBERT sentiment factor and inflation-adjusted returns. Empirical analysis shows improved risk-return efficiency over Ordinary Least Squares regression model (OLS) and the Sequential Least Squares Programming (SLSQP) under structural volatility and uncertainty.