About
The same story, told across borders.
ArgusWire is a daily editorial wire that compares how outlets from different regions frame the same story. Coverage from one country alone is incomplete. Reading the same event through six perspectives is the point.
The name
In Greek mythology, Argus Panoptes — «the all-seeing» — was a hundred-eyed giant tasked with watching what others would have preferred to leave unwatched. Some of his eyes always remained awake; nothing escaped him.
The name fits the project. Where most aggregators show you one version of an event, ArgusWire keeps fourteen pairs of eyes — mainstream and independent, across seven regions — on the same story at once. The "wire" at the end is a small homage to the original news wire services, the cables that pulled distant places into a single conversation.
What ArgusWire does
Three independent jobs run on a schedule:
- Ingest pulls RSS feeds from a curated list of mainstream and independent outlets across six regions.
- Synthesise clusters articles that cover the same event, filters for global relevance, and asks a large language model to produce a neutral, analytical synthesis — with explicit divergences — using only the source material.
- Translate renders each synthesis in Spanish, Portuguese and French via a separate translation provider, with caching so unchanged stories are never re-translated.
- Publish renders the result as a static HTML edition. No JavaScript runs in your browser unless you click "see divergences", which uses native HTML
<details>.
What ArgusWire does not do
- It does not paraphrase past ten consecutive words from any source.
- It does not invent facts. Every claim must be traceable to a cited outlet.
- It does not editorialise: voice is neutral and analytical.
- It does not hide its sources. Every article links every outlet it draws from.
- It does not track readers. There are no analytics scripts.
AI disclosure
Syntheses are produced by Anthropic's Claude Sonnet (4.6) for long-form text and Claude Haiku for filtering and clustering. Translations are produced by Llama 3.3 70B via Groq. The token count for each synthesis is shown at the bottom of the page, alongside the cluster ID. Prompts are deterministic: the model is instructed to declare framing divergences explicitly when it detects them.
AI is not a journalist. ArgusWire is an editorial tool that shows you, side by side, how different regions describe the same event. The interpretation is yours.