The author shares thoughts on LLMs and AI, preparing for a vacation and reflecting on industry trends. Surveys on AI's impact in software development often overlook varied LLM usage, potentially leading to skewed results. The author expresses uncertainty about the future of programming, urging experimentation with LLMs and sharing experiences. They believe the AI boom is a bubble, similar to past technological advancements, but with potential for lasting value despite eventual bursting. The author will attend the GOTO Copenhagen conference, reconnecting with industry colleagues. LLMs are viewed as hallucination engines, emphasizing the need to verify outputs and consider multiple prompts. Software engineering may evolve to incorporate non-determinism, mirroring other engineering disciplines. The author notes that LLMs can behave like unreliable junior colleagues, providing misleading results. Finally, the author highlights the increased security risks introduced by LLMs, particularly concerning agentic browser extensions.