The paper presents the discovery and implementation of a chemical reservoir computer based on the formose reaction, a self-organizing chemical reaction network. This system demonstrates the ability to perform nonlinear classification tasks, predict the dynamics of complex systems, and achieve time-series forecasting. The formose reaction, a complex and self-organized chemical network, can process information from its environment and perform various computational tasks, including Boolean logic gates and nonlinear classification tasks, without the need for explicit engineering of individual reactions. The study shows that the formose reaction can emulate the behavior of biochemical reaction networks and forecast changes in chaotic dynamical environments. Additionally, the system can harness short-term memory to predict future environmental dynamics. The findings provide a proof of concept for the emergent computational capabilities of complex chemical reaction networks, paving the way for a new class of biomimetic information processing systems. The approach circumvents the limitations of designed (bio)molecular computers and offers potential for rapid improvements in large-scale computation and simulation of multiscale dynamical systems.The paper presents the discovery and implementation of a chemical reservoir computer based on the formose reaction, a self-organizing chemical reaction network. This system demonstrates the ability to perform nonlinear classification tasks, predict the dynamics of complex systems, and achieve time-series forecasting. The formose reaction, a complex and self-organized chemical network, can process information from its environment and perform various computational tasks, including Boolean logic gates and nonlinear classification tasks, without the need for explicit engineering of individual reactions. The study shows that the formose reaction can emulate the behavior of biochemical reaction networks and forecast changes in chaotic dynamical environments. Additionally, the system can harness short-term memory to predict future environmental dynamics. The findings provide a proof of concept for the emergent computational capabilities of complex chemical reaction networks, paving the way for a new class of biomimetic information processing systems. The approach circumvents the limitations of designed (bio)molecular computers and offers potential for rapid improvements in large-scale computation and simulation of multiscale dynamical systems.