Imagine struggling to speak, your thoughts trapped behind a wall of missing words. This is the daily reality for many individuals with aphasia, a language disorder often caused by stroke. New research is exploring how the power of large language models (LLMs), the same technology behind AI chatbots, could help bridge this communication gap. Specifically, scientists are investigating whether LLMs can be trained to complete fragmented sentences typical of Broca's aphasia, a type of aphasia characterized by broken, halting speech. The challenge? Aphasic speech is highly variable and often lacks the grammatical structure LLMs typically rely on. To overcome this, researchers are creating 'synthetic' aphasic speech to train the LLMs. They take normal speech and apply rules based on the linguistic patterns of Broca's aphasia, essentially mimicking the disorder. This synthetic data is then fed to the LLMs, teaching them to predict the missing pieces of fragmented sentences. Early results are promising. The LLMs show an ability to reconstruct synthetic aphasic sentences, especially longer ones with more context. However, when tested with real aphasic speech, the models sometimes falter, highlighting the complexity of human language and the need for further refinement. This research has significant implications for developing assistive communication technologies. Imagine an app that could predict and complete a user's thoughts in real-time, allowing them to communicate more fluently. While challenges remain, this study represents a crucial step towards harnessing AI to improve the lives of those living with aphasia.