Imagine an AI that could predict your next medical diagnosis based on your medical history. This isn't science fiction; it's the focus of exciting new research using large language models (LLMs). Researchers are developing models that can analyze patient timelines, learning from the sequence of diagnoses, medications, and procedures to anticipate future health events. This approach, exemplified by a model called Foresight 2 (FS2), goes beyond simply answering medical questions. FS2 delves into the complexities of real hospital data, learning to predict the *next* step in a patient's journey. It's trained on a massive dataset of clinical notes, learning not just medical terms, but also the context in which they appear. This allows it to understand nuanced information, like the severity of a condition or whether a diagnosis was ruled out. The results are impressive. FS2 significantly outperforms its predecessor (Foresight 1) and even beats powerful general-purpose LLMs like GPT-4 at predicting a patient's risk of developing specific disorders within the next month. This success highlights the importance of specializing AI models for healthcare tasks. While large general models have broad knowledge, smaller models trained on focused, high-quality data can achieve superior performance in specific domains. This has huge implications for building alerting systems in hospitals, where high precision is crucial to avoid “alert fatigue” among clinicians. It also opens doors to more personalized risk prediction and prognosis, potentially revolutionizing preventative care and treatment planning. However, challenges remain. Medical ontologies like SNOMED, while helpful for standardization, may not capture every clinical nuance. The model also relies on the accuracy of the tools used to extract medical concepts from text, and improvements in these tools would further enhance FS2’s performance. Future work will need to address these limitations, explore even larger datasets, and incorporate the vast landscape of medical literature to ensure these AI-powered predictions are as accurate and unbiased as possible. The potential to transform healthcare is undeniably vast, and research like this paves the way for a future where AI plays a vital role in improving patient outcomes.