Photo by cottonbro studio on Pexels

Google’s newest AI model, Gemini 3, appears to be having a bit of an identity crisis, at least when it comes to calendar awareness. Andrej Karpathy, a prominent AI researcher with early access to the model, noted that Gemini 3 struggled to accurately identify the current year. The unexpected error, which Karpathy dubbed a “model smell,” led to amusing and potentially concerning discrepancies in the model’s responses. The incident raises questions about the AI’s comprehension of basic temporal information and highlights the ongoing challenges in achieving robust and reliable AI systems.