The Thing GPT and Claude Quietly Drop in Every Conversation
Matt Maher
May 14, 2026
Current top-tier AI models struggle to retain user intent through planning phases, often dropping up to 20% of nuanced instructions. Even as models achieve near-perfect feature planning, they fail to capture the 'why' behind complex requests, suggesting that higher reasoning settings might paradoxically decrease accuracy in intent recovery.
Key insight: The 'High' reasoning mode for both GPT-5.5 and Opus-4.7 consistently outperforms 'Extra High' or 'Max' settings in intent recovery, suggesting that excessive model reasoning can sometimes degrade the retention of original user intent.
