In this video I test what “agentic coding” actually looks like in practice by building a simple productivity app in Bubble AI. The app is inspired by Sam Altman’s idea of focusing on the most important thing (instead of staying busy), so I keep the scope intentionally small: create tasks, create categories (like work/health/personal development), rank those focus areas, and then automatically prioritize tasks based on the category ranking. I walk you through how I steer the agent with a strong prompt—being explicit about logic/behavior, describing the UI (minimal, pastel blue), and giving a concrete example scenario so the AI doesn’t guess. Once Bubble generates the first version (“Pastel Focus”), I review what it built and then refine it instead of endlessly re-prompting. I simplify the UI (one place to add tasks, one consolidated task list) and show how Bubble lets me switch between prompting and hands-on control: design, workflows (logic like signup), data types (tasks, categories, user/profile), and styles (without writing HTML/CSS). I also cover practical shipping steps like running builds, fixing issues, security tests, and deploying with a click—because the real takeaway is that the interface between ideas and working software is changing, and engineers need to stay in control while moving faster.