This tutorial is a half-day tutorial to introduce attendees to the history of large language models (LLMs) leading up to the discovery of in-context learning (ICL). This includes popular techniques used by commercial service providers to instruction-tune and align LLMs prior to deployment through public APIs. Attendees will learn how to write prompts by taking advantage of ICL prompt design patterns, including trigger words, demonstrations, and chain-of thought prompting. The tutorial will also cover task decomposition, in which complex tasks are decomposed into multiple, simple prompts to improve performance.