Leon van Zyl’s recent YouTube video, “Stop Using Claude Code Like This (Use Sub-Agents Instead),” dives into the intricacies of using Claude Code’s subagents and background agents to streamline coding tasks. As the video unfolds, Leon effectively engages users by demonstrating the parallel implementation of subagents to tackle complex tasks without clogging the main conversation thread. He illustrates the use of built-in subagents like Bash and Explorer agents and skillfully guides viewers on creating custom agents to optimize workflows. Leon emphasizes the importance of preserving the main context window by leveraging subagents to manage distinct tasks, thereby maintaining effective operation and avoiding common pitfalls of agent-based coding. He convincingly demonstrates how each agent functions in its sphere with separate context windows, yielding qualitative results without consuming the central conversation’s context. While praising the strategic insight provided, some arguments could benefit from a deeper exploration of potential challenges when scaling subagent deployment on vast projects. Despite these minor areas for improvement, his overall presentation redoubles confidence in subagent utility, shedding light on their potential nestled amidst the landscape of agentic coding. The instructional content, presented dynamically, bridges technical intricacies with relatable, practical appraisal, making it an engaging experience for both novice and experienced developers.