About Amalgam News

Amalgam News synthesizes coverage from multiple named news outlets into a single, clearly sourced article, alongside a measured look at how readers and the public are reacting on sites like Reddit, Mastodon, and Hacker News. Every article published here has been reviewed and approved by a member of our editorial team before it goes live.

How a story gets selected

We monitor a curated set of established, named news outlets through their RSS feeds. When several different outlets begin reporting on the same underlying event, our system groups those reports together. Only stories covered by at least two independent outlets are considered for publication: single-source reports don't meet our threshold for cross-verified coverage and are discarded automatically.

Stories that are very thinly sourced (fewer than three independent outlets, or very little total source material) are not shown advertisements and are excluded from search-engine indexing until the coverage develops further. We'd rather show fewer articles of genuine depth than pad the site with thin takes on breaking news.

How an article is drafted

Once a story group clears our sourcing threshold, an AI language model reads the grouped reporting alongside 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, synthesising and paraphrasing, never copying sentences verbatim from any one outlet
• Draw on all of the grouped sources rather than mainly retelling one of them
• Attribute every factual claim to the outlet that reported it
• Use careful, hedged language for allegations, charges, and unresolved legal matters
• Never assert guilt, motive, or intent as established fact

Each draft also includes editorial additions that go beyond what any single source provides: a "Why this matters" section placing the story in broader context; a "What's confirmed vs what isn't" section comparing what multiple independent sources agree on against what is still uncertain or unverified; and where relevant, a brief factual background note on the main entity or topic involved. These additions are generated from the available source material and general background knowledge, and are checked by editors before publication.

Automated quality checks

Every draft is run through a series of automated checks before it reaches editorial review:

• Source similarity: the draft must not copy phrases verbatim from any source, and its content must draw on all grouped outlets rather than leaning heavily on one
• Editorial tone: the draft must not use magazine-style language, assert interpretations as facts, or end with a moral or grand conclusion
• Allegation language: for stories involving crime, accusations, or active legal proceedings, every claim of wrongdoing must be attributed to a named source and expressed in hedged language
• Sensitive information: stories involving health, mental health, sexual orientation, religion, immigration status, or family court matters are flagged for additional editorial scrutiny
• Private individuals: stories naming a private individual in connection with a serious allegation are held unless corroborated by a minimum number of independent outlets
• Spelling and accuracy: an automated spell-check corrects obvious typos without touching names or specialist terms

Drafts that fail any of these checks are automatically rejected and the pipeline generates a fresh attempt. Only drafts that pass all checks are submitted for human editorial review.

Human editorial review

Every article, without exception, is reviewed and approved by a member of our editorial team before publication. Editors read the draft, check it against the original source reporting, and confirm that the automated quality checks reflect the article's actual content. No article is ever published automatically: human approval is required in every case.

Stories with specific editorial or legal flags — allegations against named individuals, sensitive personal information, active court cases, and similar — receive individual line-by-line review. All other articles go through a structured batch editorial review process. In both cases, a human editor must actively approve the article before it goes live.

Editors can reject articles outright, request a fresh draft, dismiss a specific flag that doesn't apply to the final text, or approve the article as written. An article that a reviewer is not comfortable approving is not published, regardless of how many automated checks it passed.

What counts as fact, and what doesn't

Only the named news outlets' reporting is treated as a basis for factual claims in the narrative voice. Public social media reaction is used solely to describe the public mood and to select genuinely quotable responses: it is never used to introduce a fact, identify a suspect, attribute a motive, or establish that an event occurred. If something is only being discussed on social media and no named outlet has reported it, it will not appear in our articles.

Where coverage involves an accusation, a criminal charge, or an active legal case, our automated checks and editorial review together require attributed, hedged language ("prosecutors allege", "according to police", "the claim has not been tested in court") rather than statements of guilt or wrongdoing as fact. A notice appears on the article itself explaining this.

What this site is not

This is not investigative journalism. Nothing here is independently verified beyond what the cited outlets have already reported, and no original reporting, interviews, or primary fact-checking is carried out. The public sentiment figures shown on each article are a snapshot of a small sample of public posts, not a representative survey, and we say so plainly when the sample is too small to be meaningful.

Corrections and takedowns

We are responsible for everything published here exactly as if a person had written it. If you believe an article contains a factual error, is unfair or misleading, or names you or someone else in a way that causes harm, please use the correction or takedown links shown on each article, or see our Corrections & Takedown Policy. We review urgent requests promptly. For anything else, use our contact page.