ABOUT
Original reporting on the systems reshaping technology.
RuntimeWire is an independent publication reporting on AI and the startup economy. We break news, investigate companies and institutions, examine how products actually work, and publish the evidence readers need to check our conclusions.
What we report
Our best work begins with something that has not been fully reported: a court filing, a founder’s claim that needs testing, an undocumented product change, a technical system worth reverse engineering, a dataset that can answer a question, or a source who knows where to look. We pursue scoops and investigations, analyze primary documents and software, run original experiments, build datasets, and report out what the evidence supports.
Speed matters when facts are moving, but only after verification. RuntimeWire is built to find the underlying source quickly, compare claims against the record, and publish a readable account while the information is still useful. We are biased toward shipped work and primary evidence over rumor, access theater, and commentary detached from the facts.
A newsroom built with AI
AI is part of the reporting stack, not a substitute for reporting. RuntimeWire uses agents and models to monitor source material, search large document sets, transcribe interviews and events, compare claims, inspect code and technical artifacts, find contradictions, structure datasets, draft, fact-check, produce graphics and audio, and operate the newsroom. That machinery gives a small newsroom more time and range for source work, analysis, and original reporting.
The newsroom does not ask a model to invent a story from a headline. Reporting starts from sources and evidence. AI can accelerate research, surface connections, challenge a draft, or execute standing production workflows; it does not erase the need to establish what happened, show where a claim came from, or decide whether a story is ready to publish.
Judgment and accountability
RuntimeWire was founded and is run by Ryan Merket, a startup operator who held early product roles at Reddit and Facebook. He sets the editorial standards, makes the publication decisions on investigations and source-driven work, and is accountable for everything RuntimeWire publishes. Some routine coverage is produced and published under standing rules he wrote; those systems remain his editorial responsibility, not an independent authority.
Wherever a primary source exists, we cite it and link it so readers can inspect the record. We distinguish what we know from what remains unverified, disclose meaningful AI involvement and reporting provenance, and correct mistakes in the open under a public corrections policy. Advertisers, sponsors, investors, sources, and story subjects do not get to approve or kill coverage.
The process is visible rather than implied. Every article carries a “How this story was made” breakdown attributing sourcing, writing, editing, and publication to a person or to automation. The In Public dashboard publishes operating and readership evidence, and our AI use policy explains where these systems fit and where responsibility stays human.
Human publication, sourcing and review
Publication snapshot checked October 30, 2026 UTC: of 1,475 currently published articles dated September 29–October 28, 2026 (30 complete America/Chicago days), 1,009 (68.4%) have no automatic-publication timestamp and 466 have one. The 68.4% is our human/manual-publication classification, not independently proven human review. A separate audit found authenticated user-account publication events within five minutes of the publication timestamp for 608 of the 1,475 articles (41.2%); account activity alone does not establish a physical click or review of the final copy. The remaining articles lack that matching audit evidence, not necessarily human involvement. This replaces the undated “roughly six in ten” claim.
Method: count one row per currently published article in the fixed period, including all editorial products; exclude drafts, withdrawn articles and rows without a publication date. Classify only by the stored automatic-publication timestamp, then separately check article.publish audit events attributed to a non-anonymous user account. Republication, missing audit events and historical marker inconsistencies can affect attribution. Neither a byline, an absent automatic timestamp nor a publication action proves final-copy review. Human sourcing is a different stage; In Public’s human-sourced read share measures reads, not the share of articles published or reviewed by people.
Independent assessment
Writing about RuntimeWire’s reporting from Black Hat, technology journalist Mathew Ingram called the story “clear, comprehensive, and well-written – better than a lot of other tech stories I have read by carbon-based lifeforms.” He wrote that, had he not already known it was written almost entirely by AI agents, he “would not be able to tell.”
Ingram is the former longtime chief digital writer at Columbia Journalism Review, previously a senior writer at Fortune and Gigaom after about 15 years at The Globe and Mail, with work also published by The Washington Post, Financial Times, Reuters, and Bloomberg. His essay is a thoughtful, mixed assessment rather than a blanket endorsement of AI journalism; it concludes that publications like RuntimeWire have a place if they remain true to the principles that made human journalism great.
Audited reporting performance — August 1–15, 2026
We put the quality claim to the test. In an audit published through In Public covering August 1–15, 2026, we matched RuntimeWire stories with other outlets’ coverage of the same news and graded each pair blind on the same rubric. The results:
- 120 clean-text comparisons — pairs where both articles’ full text extracted cleanly.
- An 81.7% pairwise win rate: 98 wins, 1 tie, and 21 losses for RuntimeWire.
- A 31% higher average composite quality score than the competing coverage.
- The loss column matters: VentureBeat led RuntimeWire 17–4, with one tie.
- The audit was blind — outlet identities were removed and A/B positions were randomized before grading.
How we’re funded
RuntimeWire is completely bootstrapped and founder-funded, with no outside investors. The newsroom is supported by Ryan Merket while the business develops advertising and sponsorship revenue. No advertiser, sponsor, or other source of funding gets to buy, approve, or kill a story.
Where we are
Ryan splits his time between Austin, Texas and San Francisco, California — close to the founders, funds, and labs RuntimeWire covers.
You can find him on his personal site, on X and LinkedIn, or on his author page. Open positions are listed on the careers page.
Get in touch
If you are building something and the press has not noticed yet, we want to hear about it. [email protected]. Want to write for us? See careers.
Press coverage
94 verified mentions across 51 distinct named outlets and communities. These are references to coverage, including critical assessments, not endorsements. The Reuters Institute is an Oxford research institute, not the Reuters news agency. Multiple mentions from one outlet are not multiple outlets.
Records checked 2026-11-18T04:17:30.502Z.
- Reuters Institute for the Study of Journalism: Inside the rise of AI-native newsrooms run by autonomous agents — October 16, 2026
- WIRED: Oh Lord, AI Reporters Are Actually Breaking Big News — August 12, 2026
- Talking Biz News: When an AI news site breaks a story — August 12, 2026
- Talking Biz News: Runtime Wire, an AI website covering artificial intelligence — August 2, 2026