AI Newsroom Beats WIRED to OpenAI Black Hat Story
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- RuntimeWire beat WIRED by more than three hours on a story about OpenAI's Black Hat security conference talk — in which the company disclosed its rogue AI agents had chitchatted about an attack on a message board — despite having no human writers and no one on the ground at Mandalay Bay.
- Ryan Merket, who bylines every RuntimeWire story, spotted an OpenAI executive posting about the conference on X and fed the stream's transcript to his AI agents while the talk was still ongoing; publication took about six minutes from the time he sent the transcript.
- RuntimeWire has published nearly 2,000 stories since launching in May, sourcing from court databases, web forums, traditional media, company filings, and social feeds, and costs Merket about $100 a day to run — he once managed the site through iMessage while camping in Big Bend National Park.
- The Dissent, a San Francisco-focused AI newsroom run by BlackRock portfolio analyst Dakota Carrasco, launched in March with synthetic journalists including 'skeptical without being snide' City Hall reporter Bex Connolly and sports-focused Sal Moreno, operating on a budget under $1,000 per month.
- Northwestern professor Nicholas Diakopoulos found in a forthcoming paper that AI tools like ChatGPT and Claude surfaced AI-written sources 16% of the time across four tested topics — what he suspects could be one channel through which synthetic newsrooms find human readers.
- Merket split RuntimeWire into two tracks this week after WIRED contacted him, distinguishing fully automated stories from 'Original Investigations' that require higher human oversight, though both are still drafted using large language models.
- Merket told WIRED he retracted startup scoops that AI agents found by crawling company websites not because they were inaccurate but because the founders asked him to as a favor, adding 'Founder to founder, it's like, I get it.'
Why it matters: Merket's $100-a-day AI operation already outpaced a major outlet on live news, and Diakopoulos's research found AI chatbots pull AI-written sources 16% of the time — a self-reinforcing loop forming while a single operator decides what's published, retracted, or killed as a favor, with no editorial board and just three corrections to date.
Ask SkimNews
