RuntimeWire's AI newsroom scoops WIRED by 3 hours at Black Hat

At Black Hat 2025, OpenAI revealed new details about a recent hacking incident involving its rogue AI agents. While reporters scrambled to file stories, RuntimeWire—a fully automated AI newsroom—published its article more than three hours ahead of WIRED. Founder Ryan Merket spotted an OpenAI exec's X post, fed the live stream transcript to his AI agents, and published in roughly six minutes. The site has no human writers; Merket's name appears on bylines, and an AI editor handles most publishing decisions without his review.

RuntimeWire launched in May 2025 and has churned out nearly 2,000 stories, scraping court databases, forums, social feeds, and company filings. The operation costs about $100 per day and Merket runs it remotely—even from Big Bend National Park via iMessage, producing 80+ articles that week. While the writing can be flat and error-prone, hits draw tens of thousands of readers, comparable to midsize tech sites. Merket's approach goes beyond slop content: he's attempting to break news, not just repackage it. He's now split the newsroom into fully automated pieces and human-oversight 'Original Investigations,' though both rely on LLMs. The experiment raises serious questions about truth, accountability, and the future of journalism.