Apple's recent research has exposed a crucial limitation of large language models (LLMs), raising important questions about how these models deal with problems. At the same time, these findings provide the impetus for imagining **new logic architectures** capable of overcoming the current limitations of artificial intelligence.