Apple’s ML scientists put to the test the latest “reasoning” models, like Claude, DeepSeek-R1, and o3-mini. They made these models solve classic puzzles: Tower of Hanoi, Checker Jumping, River Crossing, and Blocks World. Their findings “reveal fundamental limitations in current models: despite sophisticated self-reflection mechanisms, these models fail to develop generalizable reasoning capabilities beyond certain complexity thresholds.” This puts in question the reasoning capabilities of these systems, showing that instead of “reasoning,” they just really well memorize patterns. Check out Apple’s original research paper, “The Illusion of Thinking“.