A pipeline, not a chatbot.
Five stages turn the open web's firehose into one trustworthy page — and a hard rule sits at the center: nothing publishes without at least two independent sources agreeing it's a story.
Harvest the open web, in real time
We read many trusted outlets in parallel — homepages and RSS/Atom feeds — across English and Spanish. A strict freshness window keeps only current stories, so stale archive content never floats to the top.
Tag the who, what and where
Each article is tagged with the people, organizations and places it mentions, plus its topic and language, and condensed into a short summary the rest of the pipeline can reason over.
Group by meaning, not keywords
Articles are grouped by semantic similarity plus shared entities — not matching words — so different outlets covering the same event land together even when they phrase it nothing alike.
One balanced, cited page
Once a cluster has at least two distinct sources, we write a single balanced, source-traceable page. Every claim carries a citation back to the reporting it came from.
Drop anything that can't be traced
A quality-assurance pass removes any claim that can't be tied back to a real source, and validates each page before it goes live.
The two-source rule
Single-source rumors never become an InkBytes page. If only one outlet has it, we wait until it's independently corroborated.
What we don't do
No original reporting. No opinion dressed as fact. No engagement-maximizing feed. We aggregate and synthesize — and we say so.
See the pipeline's output.
Open any event and follow each claim back to the reporting it came from.
Start reading — $9/mo