Foundation Models (FMs) like GPT-4 have spurred the development of FMware, a new class of software crafted with novel roles, assets, and paradigms. FMware finds extensive use in software engineering (SE) research for tasks such as test generation and in commercial products like GitHub Copilot, despite the inherent unpredictability of FMs. Our tutorial will explore cutting-edge research and practical approaches in crafting FMware. We emphasize a software engineering perspective over artificial intelligence (AI) to focus on SE challenges and opportunities, avoiding in-depth discussions of AI and mathematics unless they are essential.