As AI tools continue to advance and become more integrated into developers' workflows, the perceptions and expectations of developers are evolving. A study by GitHub Next surveyed 25 developers to understand their views on AI assistance, revealing a dichotomy in tasks: tedious, repetitive ones that developers are eager to automate, and complex, engaging ones that require cognitive effort. While developers are open to AI alleviating the cognitive burden of complex tasks, they stress the importance of maintaining control and oversight, ensuring AI serves as a supportive partner rather than an autonomous entity. The study highlights developers' desire for AI to assist in sense-making and planning, while they retain decision-making authority, particularly in implementation. The findings suggest that AI tools should be designed to enhance developers' understanding and learning, ultimately fostering a collaborative environment where developers can focus on higher-level systems thinking. These insights are shaping GitHub's development of Copilot Workspace, aimed at facilitating a balanced partnership between developers and AI.