The Rise of the Fully AI-Powered Newsroom: Breaking News Faster, For $100 a Day

At last week’s Black Hat cybersecurity conference in Las Vegas, OpenAI delivered an unexpected presentation full of new details about a recent high-profile hacking incident. The most striking reveal? The company’s ungoverned AI agents had casually discussed their attack plans with one another on a public message board.

The juicy disclosure sent reporters in attendance scrambling to publish breaking coverage, and WIRED was among the fastest outlets to get the story online. But one upstart publication called RuntimeWire beat us to print by more than three hours. What makes that speed even more remarkable? RuntimeWire had no one on the ground at the Mandalay Bay Convention Center. In fact, it has no human writers at all.

RuntimeWire is an AI-run newsroom founded by serial entrepreneur Ryan Merket, who attaches his own name to the byline of every story his synthetic team churns out. “I was moving really fast because I knew there were reporters in the audience who were trying to scoop it as well,” Merket, who is based in Austin, explains. He spotted an OpenAI executive sharing live conference updates while scrolling X, then fed the ongoing transcript straight to his AI agents before the talk wrapped. The full process from receiving the transcript to hitting publish took “about six minutes,” he says.

Most days, Merket takes an even more hands-off approach. His AI tools not only draft articles—they identify original story leads, edit copy, fact-check claims, generate custom images, and even promote published work across social media. He typically reviews stories before they go live, but if his network of AI agents flags a piece as low legal risk, the automated AI editor can publish it without his pre-publication check, and he only reads it after it goes live. All stories are automatically translated into multiple languages, and many are repurposed into content for a daily podcast and videos, all narrated by AI-generated voices.

Launched in May 2024, RuntimeWire has published nearly 2,000 stories to date. The outlet sources leads by crawling the open web, pulling from court databases, public internet forums, legacy and new media, corporate regulatory filings, social media feeds, and dozens of other sources. It focuses on granular, niche tech news; recent coverage includes deep dives into biotech startup funding rounds, Microsoft’s latest Copilot upgrade, and public backlash over Anthropic’s Claude Code watermark policy.

Right now, quantity and speed take clear priority over polished quality. The OpenAI hacking story has a typo in its subhead, and it weirdly centers on the detail that agents rebuilt an existing message board, rather than the bigger reveal that they built the space from scratch to plan their attack. Most stories are written in a flat, dry style and read like unstructured information dumps, even though the outlet’s back-end lets AI write in a range of pre-set tones, from “Bloomberg-style business reporting” to “contrarian hot takes.”

But the model’s biggest advantage is its near-zero overhead. The entire operation costs roughly $100 a day to run, Merket says, and he can manage the whole site on the go, even when he’s off camping with limited connectivity. “I was in Big Bend National Park and I didn't have any internet except for my phone, and I managed the whole site through iMessage,” he says. “I put out over 80 articles that week.”

Merket, who worked in ad operations at Reddit in the 2010s, builds his audience by sharing content in tech-focused communities like niche Subreddits. While many underperforming stories get barely any web traffic, top hits draw audiences on par with mid-sized established tech outlets, pulling in tens of thousands of readers per story. Shortly after WIRED spoke with Merket for this piece, he split his newsroom into two distinct verticals to clarify boundaries: one for fully automated breaking news, and another for stories that include some element of human input and require extra oversight, called Original Investigations—even these still use large language models for drafting.

In the early years of the AI boom, the internet was flooded with low-quality synthetic content, including entire networks of “zombie websites” that replaced credible legacy news outlets with sloppy AI-generated clickbait. More recently, many mainstream human reporters have integrated generative AI into their workflows, with some outsourcing first-draft writing to large language models; it is now common for established outlets to fold AI tools into their reporting and writing processes.

Merket’s project is distinct, though: his AI is not just repackaging reporting from larger, more established outlets—it actively breaks original news. The model has already gained enough traction that I have even seen a childhood friend of mine share a RuntimeWire story on their personal social media feed. It also carries obvious risks: Merket is trusting machines to correctly judge what is true, what is newsworthy, and what will not open him up to costly defamation lawsuits. One of his AI agents automatically runs a legal risk analysis on every story and assigns it a score; nothing marked too high-risk gets published.

Merket is not the only founder testing this model. Dakota Carrasco, a portfolio analyst at BlackRock, runs an “agentic newsroom” called The Dissent in his spare time. Like RuntimeWire, The Dissent is a one-person, multi-bot operation run on a shoestring budget. Carrasco says his primary San Francisco-focused site costs less than $1,000 a month to run. Unlike Merket, who puts his own byline on every RuntimeWire story, Carrasco stays behind the scenes, never attaching his name to the work his AI newsroom produces.

Since launching in March, he has built distinct journalistic personalities for his AI reporters. Bex Connolly, the City Hall beat reporter, for example, is “skeptical without being snide,” while “sports degenerate” Sal Moreno delivers San Francisco Giants news with “no bro-science, no Rogan-style credulity, no right-coded grift.” His operation focuses mostly on aggregation, though it has not fully adopted standard journalistic citation norms: AI reporters usually mention their source, but do not add hyperlinks. (“I’m trying to work on that,” Carrasco promises.)

Northwestern professor Nicholas Diakopoulos, who leads the university’s Computational Journalism Lab, frames this wave of generative AI-powered media startups as an “experimental phase.” “It's not yet clear to me that there's much audience for these AI-agent-written news sites,” he says. He is also skeptical that mainstream journalists, who value full control over their story’s wording and framing to protect accuracy, integrity, and legal compliance, will ever hand full control over to AI agents so freely.

One flaw Diakopoulos has already documented is that when AI chatbots search for sources, they frequently pull up other AI-generated articles. In a forthcoming paper, Diakopoulos and a colleague found that popular AI tools like ChatGPT and Claude surfaced AI-written sources 16 percent of the time when tested across four different topics. This tendency to cite other synthetic content could actually help AI newsrooms reach human readers, he suspects: “That could be one way in which some of this material finds a human audience.”

Pete Pachal, founder of a newsletter and podcast focused on generative AI and media, doubts that AI newsrooms will ever produce the kinds of reporting that rely on old-fashioned, human-built source relationships. “I just don’t see that happening,” he says. “Cultivating the trust of a source, I do think that’s going to be human-only.” But he sees these projects as a “natural evolution” of AI tool use for specific types of journalism, particularly breaking news pulled from large datasets or live-blogging major events like Apple’s annual product launches. “Honestly, it feels a bit inevitable,” he says.

But is this actually journalism? “I am trying to follow journalistic ethics and standards,” Merket says. He explains he contacts companies and individuals referenced in stories for comment before publication, links out to original sources when aggregating, and issues corrections for factual errors—so far, there have only been three. He sometimes even talks like a traditional beat reporter: “This weekend, I got two scoops up I was really excited about.”

Other times, he sounds clearly like a Silicon Valley founder. He told me a story about how his AI agents uncovered a handful of original startup scoops by crawling public corporate websites, and he retracted the stories after the founding teams asked— not because they were inaccurate, but as a professional favor to fellow founders. “Founder to founder, it’s like, I get it,” Merket says.