About Amalgam News
Amalgam News is an automated news site. It reads a wide set of named news outlets, finds stories that several of them are covering at once, and uses an AI model to write a single article that synthesizes their reporting, alongside a look at how the public is reacting on sites like Reddit, Mastodon, and Hacker News. No human writer or reporter is involved in drafting the text you read.
How a story gets written
A background process continuously fetches articles from a large list of mainstream news RSS feeds. When several different outlets are reporting on the same underlying event, those reports are grouped together. An AI model then reads that group, in addition to a sample of genuine public reaction posts, and drafts a headline, a summary, and a full article.
The model is instructed to write in its own words rather than copying sentences from any one outlet, and to draw on all of the grouped sources rather than mainly retelling one of them. Every published article lists exactly which outlets it was synthesized from, with links to the original reporting, so you can always go and check.
What counts as fact, and what doesn't
Only the named news outlets' reporting is treated as a source of facts. Public reaction (social media posts and comments) is used only to describe the mood and to pick out genuinely quotable reactions; it is never allowed to introduce a fact, a suspect, a motive, or an accusation that isn't also reported by a named outlet. If something is only being said on social media and no news outlet is reporting it, this site won't say it happened.
Where coverage involves an accusation, a criminal charge, or an active legal case, the model is instructed to use careful, attributed language ("prosecutors allege", "according to police") rather than stating guilt as fact, and the article carries a notice explaining that.
Automated checks before publication
Every draft is run through a set of automated checks before it goes live: whether it copied too closely from a source, whether it actually drew on all the grouped sources instead of leaning on just one, and, for sensitive stories, whether it stated an accusation as established fact instead of attributing it. A separate classifier looks at every story for legal and editorial risk, for example whether it names a private individual, a child, or someone's health, and stories that meet that bar are held for a human to review and approve before they're allowed to publish, rather than going live automatically.
Articles are also passed through a conservative spell-check pass after drafting, which fixes obvious, unambiguous typos without touching names or terms that are unfamiliar to a standard dictionary, so a real person's name is never mistaken for a misspelling.
What this site is not
This is not investigative journalism. Nothing here is independently verified beyond what the cited outlets already reported, and no original reporting, interviews, or fact-checking is done. The "public sentiment" figures shown on each article are a snapshot of a small sample of public posts, not a survey, and the site says so plainly when that sample is too small to mean much.
Who's responsible
An AI model writing the text doesn't change who's accountable for it: we are responsible for everything published here, including headlines, exactly as if a person had written it. See the Corrections & Takedown Policy for how to request a correction or an urgent takedown.