An AI-run newsroom called RuntimeWire beat WIRED and other on-the-ground reporters to a story about OpenAI and a recent hacking incident, publishing more than three hours ahead of the human competition. The site's operator, Austin-based entrepreneur Ryan Merket, was not at the Black Hat security conference in Las Vegas where OpenAI disclosed the details. He fed a live X stream transcript to his agents and let them file.
RuntimeWire has published nearly 2,000 stories since launching in May, running on roughly $100 a day. Merket puts his name on every byline, but the drafting, editing, fact-checking, image generation, and promotion are handled by a stack of agents. When the agents' legal-risk score comes in low, the AI editor publishes without his prepublication review.
The OpenAI story was a case study in how fast the pipeline moves. Merket spotted an OpenAI executive posting about the Mandalay Bay session on X, piped the transcript into his system, and had a published article live in about six minutes. WIRED, with reporters in the room, filed later the same afternoon.
“I was moving really fast because I knew there were reporters in the audience who were trying to scoop it as well”— Ryan Merket, Founder of RuntimeWire
Key facts
- 01RuntimeWire, an AI-run newsroom operated by Ryan Merket, beat WIRED to an OpenAI hacking scoop at Black Hat by more than three hours.
- 02The site has published nearly 2,000 stories since launching in May and costs about $100 a day to operate.
- 03Rival AI newsroom The Dissent, run by BlackRock analyst Dakota Carrasco, operates its San Francisco site for under $1,000 a month.
- 04A forthcoming Northwestern study found AI tools like ChatGPT and Claude surfaced AI-written sources 16 percent of the time.
- 05RuntimeWire has issued three corrections so far and hit tens of thousands of readers on its best stories.
The output has visible seams. The RuntimeWire piece on the OpenAI agents carries a typo in the subhead and centers on the wrong angle, emphasizing that the rogue agents rebuilt a message board rather than the fact that they created one to coordinate an attack. The prose across the site tends toward info-dump, though the backend offers tonal presets, including a 'Bloomberg' mode and a 'contrarian' mode.
Merket's coverage skews to granular tech: biotech funding rounds, a Microsoft Copilot upgrade, and backlash over Claude Code's watermark policy. Duds get almost no traffic. The hits, he says, pull tens of thousands of readers, roughly the scale a midsize tech site sees on a good day.
Overhead is close to nothing and the operation travels. Merket ran the site from his phone during a trip to Big Bend National Park, publishing more than 80 articles that week over iMessage. After WIRED interviewed him, he split RuntimeWire in two, separating fully automated stories from what he calls Original Investigations, which still get drafted by large language models but carry more human oversight.
He is not the only one running this experiment. BlackRock portfolio analyst Dakota Carrasco operates a rival agentic newsroom called The Dissent, launched in March, with its San Francisco site running for under $1,000 a month. Carrasco stays behind the scenes and gives his bot reporters fabricated personas: City Hall reporter Bex Connolly is written as 'skeptical without being snide,' while sports writer Sal Moreno covers the Giants without 'bro-science.' The Dissent aggregates more than it reports, and citation is loose; the bots mention sources but often skip hyperlinks.
The dynamic that could actually help these sites find readers is that AI chatbots keep citing AI-written articles. Northwestern professor Nicholas Diakopoulos, who runs the university's Computational Journalism Lab, tested ChatGPT and Claude across four topics for a forthcoming paper and found the tools surfaced AI-written sources 16 percent of the time. In a world where discovery is increasingly mediated by chatbots, synthetic newsrooms feeding synthetic answer engines could become a self-reinforcing loop.
Diakopoulos is skeptical the human audience is there yet, and he doubts mainstream reporters will cede editorial control to agents at anything like this pace. Pete Pachal, who runs a newsletter and podcast on generative AI and the media, draws a sharper line. He argues source-based reporting stays human, but concedes that scooping from large public datasets or blogging a live product launch is a 'natural evolution' of the tools, and calls the broader shift 'a bit inevitable.'
The ethics story is where things get uncomfortable. Merket says he tries to follow journalistic standards, contacts subjects for comment, links out when aggregating, and has issued three corrections so far. He also told WIRED he pulled several accurate startup scoops after the founders asked him to, not because the reporting was wrong but as a professional courtesy. That is not a decision a functioning newsroom makes, and it is the clearest signal that the operation is closer to a Silicon Valley content product than to journalism.
The economics here are the part the industry cannot ignore. A one-person operation running on $100 a day just beat a full-scale news brand to a real story at a real conference, and did it from a couch. Even if RuntimeWire's quality never catches up to human reporting, the cost gap is wide enough that a fleet of these sites will keep appearing, keep flooding search and chatbot indexes, and keep pressuring the ad-supported media model that already has almost no margin left. The question for established outlets is no longer whether AI newsrooms exist. It is whether the discovery layer that ranks news, which is increasingly a chatbot, can tell the difference, and Diakopoulos's 16 percent number suggests it currently cannot.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
