Simon Willison's experiment points to a potentially significant, if subtle, shift in software engineering practice. The value proposition here may not be the agent itself—many similar tools exist—but the demonstration of using AI to bootstrap the entire development lifecycle, from spec to tests. This could lower the barrier for creating niche, single-purpose agents, but it also arguably creates a new class of technical debt: code whose original intent and structure is opaque, even to its nominal author.
As Willison's work suggests, the future of tool-building may involve less writing from scratch and more curating and directing AI-generated outputs, a change that could reshape foundational developer skills.