A wave of recent research is fundamentally challenging how we understand large language models by exploring emotional signals and behavioral nuance. A new mechanistic study examines whether emotional cues—analogous to feelings in humans—can shape how LLMs make decisions and perform tasks. This investigation builds on the recognition that emotions profoundly influence human cognition and performance. The findings suggest that emotional signals may similarly modulate LLM behavior in meaningful ways, opening questions about whether we need to incorporate emotional awareness into AI system design for better outcomes.