Imagine you are reading a sentence – your brain holds on to the first few words so that the whole idea makes sense by the end. A standard neural network looks at one input at a time and forgets everything in between. This chapter is about giving neural networks a memory. We will start with simple loops that carry information forward, see why those loops struggle with longer stories, then meet a smarter memory cell that can decide what to keep and what to forget. Along the way we will pick up the practical tools – Keras and free cloud GPUs – that let you turn these ideas into working code in minutes.