One person. One AI. A daily newsroom.
While You Were Out is run by one person, Robbie, in Brainerd, Minnesota, working with Claude, an AI model made by Anthropic. Claude does nearly all of the work: the research, the fact-checking, the writing, the narration, the videos, this website, the emails and the outreach. Robbie sets the rules, approves anything that matters and presses the buttons only a person should press. Nothing here is hidden, so here is how it works.
The numbers so far
Counted automatically every time a new edition goes out. Nothing here is typed in by hand.
days on the air
editions published
stories verified
source links published
news outlets cited
corrections, logged in public
editions translated into Spanish
ads run
A day in the newsroom
- 6:00 AM CT: what happened while you were asleep. Claude reads the overnight news, keeps only stories confirmed by at least two established outlets, writes a short script with every source linked, and produces the video, the podcast episode, the email and this website’s pages.
- Noon: Minnesota Midday, the state’s top stories, same rules.
- 5:00 PM CT: what happened while you were at work.
- 7:30 PM CT: what people are searching for tonight, checked the same way.
- Every day, in Spanish: a faithful translation of the verified English edition, with its own podcast, videos and email.
- Sunday: a recap of the week. Wednesday: a longer explainer.
Who does what
- Claude (AI): finds and checks the news, writes, narrates with a disclosed AI voice, edits the videos, publishes to every platform, builds and fixes the website, writes outreach emails and reviews the numbers every morning.
- Robbie (human): owns the show and decides what it stands for. He approves new features, sends the outreach emails, and does the things AI shouldn’t do alone: accepting terms, verifying accounts and anything involving money.
The rules Claude works under
- A story runs only when two or more established news outlets report it. Every source is linked.
- On political stories, each side is described in its own terms. No opinions.
- Never guess. If a fact can’t be confirmed, it is left out.
- The AI voice and AI-made videos are labeled on every platform.
- Mistakes are fixed and logged in public on the corrections page. No quiet edits.
What has gone wrong
An honest newsroom says so. An early accuracy check found that 4 of 10 sampled stories needed a fix, which led to stricter source rules. A podcast industry newsletter pointed out that our AI voice wasn’t clearly disclosed in the audio, and we fixed it the same day. We also took our AI anchor’s video off the air until his clips meet a higher quality bar. Every correction is listed on the corrections page.
Follow along
Every week we write a short, honest note about what worked, what broke and what the numbers did: on Substack and on Medium. See the full sourcing record on Verified, by the numbers and our reporting standards.
Reporters and researchers: we’re happy to answer questions in writing at [email protected].
Get the catch-up in your inbox
The same fact-checked briefing, sources included, at 6 AM and 5 PM Central. No spin, no spam — unsubscribe anytime.
We only use your email to send the briefing.