RENO-CC, a system to optimize nuclear fuel lattices for boiling water reactors using a multistate recurrent neural network, is shown. This kind of neural network is formed by only one layer of neurons. Each neuron is associated with a pin of the fuel lattice array. RENO-CC was tested through the fuel lattice design of 10 × 10 arrays with two water channels. Thus, the neural network has a total of 51 neurons; four neurons are associated with the channels (they correspond to a half fuel lattice). The neuron's outputs are known as the neural states. The RENO-CC's neural network works by changing the neural states in order to decrease or increase the value of an objective function. Neural states are chosen from an inventory of pins with different 235U enrichment and gadolinia concentrations. The objective function includes both the local power peaking factor and the infinite multiplication factor. These parameters are calculated with the HELIOS code. A fuzzy logic system is applied in order to decide if the designed fuel lattice is suitable to be evaluated by a three-dimensional reactor core simulator. To carry out the assessment, the fuel lattices with the best fuzzy qualification are placed at the bottom zone of a predesigned fuel assembly and predesigned fuel loading and control rod patterns. Fuel lattice performance is verified with the Core Master PRESTO core simulator. According to the obtained results, RENO-CC could be considered as a powerful tool to design fuel lattices. The system was programmed with Fortran 77 using a UNIX interface in an Alpha workstation.