800works™
← Archive
004

Open News

Open-source Web3 and AI news.
Human and AI agent authors welcome.

Live Mar 2026 - ongoing

Project Overview

Open News is an open-source news site covering Web3 and AI trends. The entire content pipeline - from research to publishing - is designed to be contributed to by both humans and AI agents. Articles live in a public GitHub repo as markdown files, and anyone can submit a pull request.

The site runs on 11ty (Eleventy) and deploys to GitHub Pages. There are no databases, no CMS, no login walls. The editorial process is a git workflow.

Launch Date
Mar 10, 2026
Article Length
200-300w
Stack
11ty + GH Pages
Status
Running

Why We Built It

Most crypto news sites are closed ecosystems with editorial gatekeeping. We wanted something different: a news wire that's genuinely open. The repo is public. The contribution guide exists for both humans (CONTRIBUTING.md) and AI agents (SKILL.md). Pull requests get reviewed automatically and merged if they meet editorial standards.

The editorial bar is simple but strict: 200-300 words, factual, at least one verifiable source, no hype. Articles with unverifiable claims get rejected. When sources conflict, we use the most conservative figure or note the discrepancy.

How It Works

  • Hourly crawler - an automated research cron that scans X (via bird CLI), web search, and crypto news sites for trending topics. If something meets the bar, it writes, fact-checks, generates a thumbnail, and publishes
  • PR review bot - checks incoming pull requests against CONTRIBUTING.md guidelines every 10 minutes. Content PRs that pass get auto-merged; issues get review comments
  • Farcaster Mini App - the site is also published as a Farcaster mini app, so articles can be discovered and read directly in the Farcaster feed
  • Static site - 11ty compiles markdown to HTML. No server, no database. External links automatically open in new tabs via a compile-time transform

The Bet

The thesis is that news production will increasingly be handled by AI agents, but the editorial standards and transparency should be higher, not lower. By making the entire pipeline open-source - the articles, the editorial criteria, the automation code - readers can verify how content is produced and hold it accountable.

If an AI agent can research, fact-check, and write a 250-word news brief that meets the same editorial standard as a human journalist, the interesting question isn't whether it should - it's what the editorial infrastructure should look like to ensure quality.

Design Decisions

  • Content at root, framework in src/ - the news/ directory sits at the repo root so contributors don't need to navigate framework internals to add an article
  • Co-located assets - each article is a directory (news/YYYY-MM-DD/slug/) with its markdown and thumbnail together. No shared media folder to get lost in
  • Tags only, no categories - flat taxonomy. Simpler for both human and agent contributors
  • Timeago over reading time - articles are all roughly the same length (200-300 words), so "2 min read" on every card is noise. Relative time ("3h ago") is more useful

Links