Large Language Models (LLMs) have shown impressive abilities across various domains, but can they truly grasp the complexities of physics? While LLMs can generate human-like text, their performance on physics problems reveals some fundamental limitations. They often struggle due to a lack of specific physics knowledge, like key formulas and concepts. Even when provided with the necessary information, they sometimes misapply it, revealing a gap in their reasoning abilities. Researchers have introduced a new framework called Physics Reasoner to address these shortcomings. This framework augments LLMs with a comprehensive set of physics formulas and employs checklists to guide the problem-solving process. Think of it as giving the LLM a textbook and a study guide. The Physics Reasoner works in three stages: problem analysis, formula retrieval, and guided reasoning. It first breaks down the problem, then fetches relevant formulas from its knowledge base, and finally uses the checklists to ensure it applies the formulas correctly. Experiments show that Physics Reasoner significantly improves LLMs' accuracy on physics problems, especially complex ones. This suggests that augmenting LLMs with domain-specific knowledge and structured reasoning strategies is key to unlocking their full potential in scientific domains. While this research is promising, it also highlights the ongoing challenges in developing AI that can truly reason like a physicist. Building a more comprehensive formula set and refining the checklists for different physics subfields are crucial next steps. This research also raises broader questions about the role of knowledge and reasoning in AI. Can LLMs eventually learn to acquire and apply knowledge independently, or will they always require external guidance? As LLMs continue to evolve, tackling these challenges will be essential for building AI that can not only solve complex scientific problems but also contribute to new discoveries.