We optimize for friction relief in the moment, outsourcing the cognitive load of curation to systems that learn our blind spots faster than we do.
2
A mail carrier like John Ayala in Los Angeles learns his neighborhood by showing up, by friction—dogs, weather, the small human moments that build knowledge. An algorithm learns you by removal of friction, by what you tell it you want, then by what it observes you actually engaging with. One builds judgment through resistance. The other builds prediction through preference data.
3
The trap is that preference-mapping feels like personalization but functions as narrowing. Every customization layer makes the feed more comfortable and the rest of the world less visible. Within five years, no one will remember that discovery used to require stumbling. By then, the stumblings will have become someone else's product.
Google now asks you to describe what you want to see, then remembers your answer and shapes tomorrow's feed accordingly. The exchange feels like freedom, like being understood. What's actually happening is more consequential. You're training an algorithm to predict your preferences better than you can articulate them, and you're doing it voluntarily because the alternative—scrolling through noise—feels worse.