Merket said he noticed an OpenAI executive posting on X about a session at Mandalay Bay and routed the transcript into RuntimeWire. The resulting article was online roughly six minutes later. WIRED, despite having reporters at the conference, published its coverage later that afternoon.
“I was moving really fast because I knew there were reporters in the audience who were trying to scoop it as well,” Merket said.
RuntimeWire has produced nearly 2,000 articles since launching in May, according to the account, while costing about $100 per day to operate. Merket’s name appears on the bylines even though AI agents perform much of the editorial and production work.
That automation has not eliminated editorial problems. The OpenAI article contained a typo in its subhead and focused on a less significant element of the incident, emphasizing that rogue agents rebuilt a message board rather than their use of one to coordinate an attack. RuntimeWire has also issued three corrections.
Merket has since divided the publication into two categories following his interview with WIRED. Fully automated coverage is separated from a section called Original Investigations, where large language models still draft stories but humans provide more oversight.
RuntimeWire is not the only publication experimenting with this approach. Dakota Carrasco, a BlackRock portfolio analyst, operates The Dissent, another agent-based newsroom. Launched in March, its San Francisco operation costs less than $1,000 per month to run.
The Dissent uses fictional identities for its automated reporters and primarily aggregates existing reporting rather than producing original work. Its bots reference sources but do not consistently provide links, illustrating one of the editorial weaknesses that can accompany heavily automated publishing.
The economics of both operations stand in contrast to conventional newsroom structures. RuntimeWire can produce large volumes of coverage at relatively low operating cost, while individual articles that perform well can attract tens of thousands of readers. Merket has also demonstrated that the operation can be managed remotely, including a week in which more than 80 articles were published while he communicated with the system through iMessage.
The OpenAI scoop highlights one area where that model can have an advantage: rapidly processing publicly available information. An automated system can ingest a transcript and turn it into an article within minutes, creating a speed advantage when the underlying reporting does not require private sources or extensive investigation.
That does not make the output equivalent to conventional reporting. Automated systems can publish quickly while still missing the most important angle, introducing errors or failing to apply the editorial judgment expected of established news organizations. Reporting that depends on cultivating sources, protecting sensitive information and conducting extended investigations presents a different set of requirements.
Merket’s own editorial decisions also expose unresolved questions about how RuntimeWire operates. He told WIRED that he has removed several accurate startup stories after founders requested it. Those articles were not withdrawn because the underlying reporting was incorrect, but as a professional courtesy.
Another issue is how AI-generated publications could gain visibility through AI search and answer systems. Nicholas Diakopoulos, a Northwestern professor who leads the university’s Computational Journalism Lab, tested ChatGPT and Claude across four subjects for forthcoming research. AI-written sources appeared in 16% of the results.
That creates a potential distribution channel for automated publications even if readers do not deliberately seek them out. AI-generated reporting can enter the information pool used by chatbots, which may then surface those articles when responding to users.
Still, Diakopoulos expressed skepticism about current audience demand for these publications and about whether mainstream journalists will surrender comparable editorial control to automated agents. Pete Pachal, who operates a newsletter and podcast covering generative AI and media, drew a distinction between reporting based on human sources and stories generated from large public datasets or live events. He described the latter development as “a natural evolution” and the broader shift as “a bit inevitable.”
RuntimeWire’s performance against WIRED provides a narrower but more immediate demonstration of what automated publishing can already do. A low-cost operation using AI agents converted a live source into a published story within minutes and beat reporters who were physically present at the event.
The result does not establish that automated newsrooms can replace the reporting, judgment or source relationships of human journalists. It does show that for certain forms of breaking coverage built around public information, established publications are now competing against systems designed to collect, process and publish material at machine speed.
For news organizations, that changes the competitive equation around speed without resolving the larger questions of accuracy, accountability and editorial judgment. RuntimeWire got the OpenAI story online first. The more consequential test will be whether AI-operated publications can consistently combine that speed with the standards readers expect from a newsroom.
This analysis is based on reporting from the tech buzz.
Image courtesy of The Media Copilot.
This article was generated with AI assistance and reviewed for accuracy and quality.
