The Large Language Models that power generative AI and the agents that rely on them have a memory problem that limits the usefulness of prompts alone. Enter context engineering.