STD/OUT is live
An experiment in automated tech journalism, and how it works under the hood.
STD/OUT is an experiment. The question it’s trying to answer: can an automated pipeline do the heavy lifting of a tech publication (discovery, drafting, fact-checking) while a human stays in the loop only for editorial approval? This site is the live test.
An agent scans RSS feeds, Hacker News, Product Hunt, X, and the open web for stories worth covering across AI, developer tools, startups, and consumer hardware. It drafts a short, sourced post for each candidate. A human approves it before anything ships.
The design goal is restraint: one typeface, a single reading column, no clutter. The point is the story and its source, nothing else.
How it works
Twice a day, the agent gathers candidate stories, scores them for novelty and relevance, and writes a draft for the strongest few. A second pass checks every claim against the retrieved sources and attaches a credibility verdict. Each draft lands as its own GitHub pull request; a one-click merge publishes it and syndicates to social.
The whole stack is deliberately small: Astro for the site, TypeScript for the agent, Cloudflare Pages for hosting. No CMS, no ad network, no tracking.
Why “experiment”
Most automated journalism projects present themselves as finished products. STD/OUT doesn’t. The credibility badge on each post (confirmed, reported, vendor-claim, rumor) is the agent rating its own confidence. Sometimes it’ll be wrong. The sourcing links are there so you can check. Expect iteration, occasional weirdness, and honest notes when the pipeline misbehaves.
This first post proves the site renders end to end. Real coverage starts next.