AI-assisted coding is still in its infancy. This implies that we should embrace it with caution, guardrails, and realistic expectations beyond any marketing hype. In this talk, Adam presents the short and long term implications of using AI assistants, and proposes a new way forward where the focus is on improving existing code. The talk is based on Adam's recent research of more than 100k of AI refactorings performed on real-world codebases.