Why Apples Critique of AI Reasoning Is Premature

Recent debates over the reasoning capabilities of Large Reasoning Models (LRMs) have been intensified by conflicting studies from Apple and Anthropic. Apples research claims that LRMs, such as Claude-3.7 Sonnet and DeepSeek-R1, exhibit fundamental limitations in solving complex puzzles like Tower of Hanoi and River Crossing, especially as problem complexity surpasses certain thresholds, leading to an "accuracy collapse" and reduced reasoning effort at higher complexities. The study suggests that these models struggle with exact computation and consistent algorithmic reasoning, particularly in high-complexity regimes, indicating inherent limitations in their reasoning abilities