What is Signal Board?
Signal Board is a project I'm building to see what the news looks like when you read all of it at once. Every morning, an AI pipeline reads 300 sources across national, international, local, nonprofit, conservative, liberal, ethnic, community, and specialist outlets, then answers a set of questions I wrote about where stories connect across domains, where the coverage gaps are, and where people are actually cooperating.
The daily synthesis is AI-generated from a framework I built. What I think is distinctive about it, and I might be wrong, is the framing: the questions are about structural patterns in who owns, funds, and distributes the news, and what shows up when you hold that many sources next to each other that any five of them alone could never reveal.
I'm learning as I go and publishing every day because I think other people might want to see the picture too.
How it works
Every morning at 7 AM UTC, an automated pipeline fetches articles from 300 RSS feeds covering 300 unique publications. Those sources span 12 tiers: 31 national outlets, 53 international, 51 specialist, 108 local and regional newsrooms, plus podcasts, newsletters, think tanks, government feeds, and research organizations. The complete source list is published in the project's open-source repository.
An AI classification step reads each article and tags it by structural domain (economics, governance, security, climate, labor, and more) and by structural force, which is the underlying pattern driving the story rather than its surface topic. The classifier also asks a specific question about every article: where in this story are people being decent? That cooperation dimension is how Signal Board finds the stories that no algorithm optimizing for engagement would ever surface.
A second step clusters articles by structural force and analyzes how different outlet types frame the same story. A third step uses a larger AI model to write the daily editorial synthesis, the framing analysis, and the three featured sections: The Daily Thread (widest cross-tier coverage), The Daily Gap (where framing diverges most sharply), and Meanwhile (who showed up today).
The AI models are Claude by Anthropic: Haiku for classification (about $0.50 per day for roughly 1,000 articles) and Sonnet for synthesis (about $0.40 to $1.00 per day for two editorial passes). The full source list, the classification prompts, and the synthesis prompts are all in the public GitHub repository.
Source transparency
Every source card on the site includes a one-line description of who the outlet is: their ownership, funding model, and editorial focus. This is not a bias rating; it is context. All outlets have interests, and readers deserve to know what those interests are before deciding how much weight to give a particular frame.
Signal Board does not editorialize about individual outlets. It names the framing differences, describes what each outlet emphasized and what it left out, and lets the reader draw their own conclusions. If you think a source description is inaccurate or unfair, please reach out.
Corrections
Signal Board is AI-generated and published daily. If a synthesis contains a factual error, a misattribution, or a characterization you believe is inaccurate, please contact Elise directly. Corrections will be noted on the relevant daily edition and the corrected version will replace the original.
Privacy
Signal Board does not use cookies, does not track you across the web, does not sell data, and does not display advertising. There is no login, no account, and no personalization. The site is a static page rebuilt once daily and served directly to your browser with no algorithmic intermediary.