In 1976, Charles Rich and his colleagues pioneered the concept of aProgrammer's Apprentice—an interactive programming system designed toassist expert programmers in the design, coding, and maintenance oflarge and complex programs. The idea of a digital prosthesis, extendingour biological brain to seamlessly bridge the gap between our ideas andthe code we produce, has been revisited multiple times throughout theyears. For instance, the Probability/BigCode team has spent over half adecade applying machine learning to enhance efficiency for engineers,data scientists, and systems across the board.Historically, these endeavors relied on traditional (symbolic) AItechniques or early-stage neural net models, resulting in limitedsuccess in meeting the high expectations. However, the emergence ofgenerative models like GPT has transformed what was once consideredscience fiction into reality, rendering the need for custom modelarchitectures and embeddings obsolete. Consequently, many developers arenow incorporating AI-based "coPilots" into their daily programming routines.In this talk, we will explore the various ways AI has been applied atMeta, such as code search, code recommendations, and bug fixing, andrevisit these areas in the context of large language models (LLMs).Additionally, we will discuss our vision for the future of generative AIin productivity tools, pivoting from a collection of task-specific toolsdesigned for a generic user base to a single, user-specific tool capableof tackling a multitude of tasks. This transition will illuminate theongoing evolution of AI-assisted programming and its potential impact onthe developer community and beyond.