Upworthy
Upworthy is the media company Eli Pariser co-founded in March 2012 with Peter Koechley, designed to test whether substantive content — on social justice, public health, climate, civic issues — could achieve the viral reach that algorithmic platforms typically reserved for entertainment and trivia. The company became famous for its curiosity-gap headlines ("What she did next will amaze you") and for a period in 2013 reached over 87 million monthly unique visitors, making it one of the fastest-growing media properties in internet history. Upworthy represented Pariser's practical engagement with the architecture of attention he had diagnosed in The Filter Bubble: if algorithms shaped what circulated, the question became whether the shaping could be directed toward materials that served rather than degraded public discourse.
In the AI Story
The company's trajectory illustrates both the possibilities and the limits of working within algorithmically optimized media environments. The early success demonstrated that substantive content could achieve scale when packaged with the attention-capture techniques that viral media had developed. The subsequent decline — as platform algorithms evolved, as the headline style became overused, as user attention fatigue set in — demonstrated that working within the optimization logic meant accepting its volatility.
Pariser's subsequent reflection on Upworthy's history informs his analysis throughout this book. The experiment was not a failure; it was an education. It revealed that engagement optimization is not neutral regarding content — it favors certain cognitive patterns, certain emotional registers, certain forms of attention-capture — and that attempting to redirect the optimization toward better content still leaves the underlying architecture intact. The deeper work of changing the architecture itself became the focus of his later efforts, particularly through New_ Public.
The relevance to AI is direct. Upworthy tested whether working within existing algorithmic architectures could produce better outcomes. The answer turned out to be partially — for a while — under specific conditions. Applied to AI systems, the same question arises: can builders working within AI-mediated workflows produce better outcomes by adjusting their prompts and practices? The answer, similarly, is partially — for a while — under specific conditions. The deeper intervention, as with Upworthy's successor work, requires changing the architecture rather than optimizing within it.
Origin
Pariser founded Upworthy with Peter Koechley in 2012. The company was acquired by Good Worldwide in 2017 and continues to operate, though at a smaller scale than its 2013 peak. Pariser left active management to focus on New_ Public and related civic-technology initiatives.
Key Ideas
The experiment tested content-level intervention against architecture-level limits. Better content could achieve scale within algorithmic optimization, but only temporarily and only on the optimization's terms.
Viral scale revealed optimization's content preferences. What went viral told Pariser as much about the architecture as about the content.
The trajectory informed the shift to architectural work. New_ Public emerged from the recognition that content-level intervention was insufficient.
The lesson transfers to AI. Prompt-level intervention in AI workflows faces analogous limits to headline-level intervention in media workflows.
Appears in the Orange Pill Cycle
Further reading
- Eli Pariser, The Filter Bubble (Penguin Press, 2011)
- Derek Thompson, Hit Makers (Penguin Press, 2017)
- Nicholas Lemann, "Two Peas in a Pod" (Columbia Journalism Review, 2014)