The Tyranny of the Marginal User
Why consumer software gets worse over time: companies optimize for the next user who might leave (the "marginal user") rather than existing users who derive real value, creating a race to the bottom where apps become zombified infinite scroll machines.
Read Original Summary used for search
TLDR
• OKCupid went from sophisticated personality matching (eerily accurate 95%+ match scores) to just another Tinder clone—a microcosm of how consumer software systematically degrades despite billions in R&D
• The economic mechanism: flat-fee/ad models + network effects mean companies care about marginal user retention, not existing user value, so they optimize for "Marl"—the next user who might churn
• "Marl" has goldfish attention span, zero tolerance for UI complexity, and only scrolls—features that increase agency (preferences, settings) actually decrease DAUs in A/B tests because they interrupt the scroll
• We've all been "Marl" (scrolling in bed, at airport, avoiding painful thoughts)—the entire digital economy is designed to exploit this diminished state rather than serve our intentional, creative selves
• Tools that enhance human agency remain hobbyist projects used by nerds; if they get successful, Marl-serving companies acquire and kill them
In Detail
The piece opens with a concrete example: OKCupid in 2016 had long essays, hundreds of personality questions, and match scores that were "eerily good"—but it devolved into another swipe-based Tinder clone. This isn't isolated: Google Search has decayed, most consumer software trends toward minimal agency and infinite scroll, while Reddit and Craigslist remain useful precisely because they're frozen in time. The puzzle: how does software get worse despite massive R&D budgets and AI progress?
The answer is the "Tyranny of the Marginal User." Companies optimize for DAUs (Daily Active Users), but since most products charge flat fees or are ad-supported, they don't care about existing users—they care about the marginal user, the next person who might churn. The author personifies this as "Marl": 1.3 second attention span, zero UI complexity tolerance, only capable of zombie scrolling. Features that increase user agency—preferences, settings, "see less" buttons—actually DECREASE DAUs in A/B tests because they interrupt the scroll with a few pixels that could have been a triggering headline. So these features get killed, designers don't get promoted, and apps devolve into engagement traps.
Critically, "Marl" isn't just a person—it's a state of mind we all enter when scrolling half-consciously in bed or reflexively opening our phones to avoid discomfort. The world's most brilliant people, armed with unlimited capital and god-like computers, spend their lives serving this diminished state. Meanwhile, tools that enhance human agency remain hobbyist projects. If they get too successful, Marl-serving companies flush with ad revenue or VC money acquire and kill them. The result: systematic degradation of human agency in digital products, driven by rational economic incentives.