Real news, in real time.

NUZU groups headlines from 200+ trusted publishers into story clusters, so you can see how the same event is being reported across many different sources — all on one screen.

Overview — How it works

Our Mission

NUZU News exists to give readers a clear, honest, multi-source view of what is happening in the world, without the filter of any single publisher's editorial line. We believe people make better decisions when they can see how the same event is being reported from many different angles at once — rather than through the lens of a single outlet or algorithmic recommendation feed.

We are not a publisher. We write nothing. We host no articles. We simply gather, cluster, and present — and we do it with as much transparency about our methods and sources as we can manage.

How We Curate News

Our curation process is deliberately mechanical. Human editors set the rules; software applies them every hour.

1. Sources are pre-approved. We maintain a hand-curated list of 200+ RSS feeds from publishers we have reviewed and tiered (see Source Standards). Nothing appears on NUZU from a source that hasn't been vetted.
2. The bot fetches hourly. An automated build pulls fresh headlines from every approved feed, then filters out opinion, podcasts, live-blogs, and content flagged as sponsored.
3. Clustering. Headlines are normalized, de-duplicated, and grouped by topical similarity so one event becomes one cluster.
4. Ranking. Each section is sorted by recency and the breadth of coverage — an event covered by ten flagship outlets ranks above one covered by a single partisan blog.
5. Display. Every headline shows its publisher, its reliability tier, and a link directly to the source. Nothing is paraphrased, rewritten, or hosted on our servers.

The full logic lives in our public bot.py script on GitHub. If you want to see exactly how a story reached your screen, read the code.

Source Standards

We choose sources by the same criteria regardless of ideology, geography, or medium:

• Editorial accountability. The outlet publishes bylines, runs corrections, and has a real newsroom (or clearly labeled opinion/analysis staff).
• Fact-based reporting. Primary-source reporting, documented claims, and separation of news from opinion.
• Independence from the subject. We avoid promotional outlets controlled by the entity they're reporting on.
• Reach and relevance. We include wires, flagship national papers, regional press, and specialty outlets so readers see more than one perspective on any story.

Every approved source is assigned to a reliability tier so readers can weigh what they're looking at.

Reliability Tiers

News Values & Principles

The rules we apply to ourselves and the feed:

• Accuracy over speed. If a story is contested, we want multiple sources in the cluster before promoting it.
• Transparency. Every headline names its publisher and its reliability tier. No anonymous "sources say" framing from us.
• Pluralism. We deliberately mix wires, flagship outlets, and regional press. Readers should never see just one voice on a major story.
• Non-endorsement. Inclusion of a source is not an endorsement of that source's accuracy on any given story.
• Respect for publishers. We link, we don't copy. If a publisher asks us to remove them, we do (see Corrections Policy).
• Respect for readers. No tracking, no account required, no algorithmic personalization, no dark patterns.

AI Use Disclosure

In the interest of honesty, here is exactly where and how we use AI and automation:

• Headline clustering. A deterministic, rule-based algorithm (not a large language model) groups similar headlines. It runs hourly and its logic is fully visible in our public source code.
• Source-tier classification. Tiers are assigned by human review, not by an AI model.
• Site generation. The homepage, RSS feed, and JSON feed are generated by a Python script that runs on a schedule. There is no AI-generated prose anywhere on the site.
• Support tasks. We sometimes use AI coding assistants to help write and review the build scripts. Any code they produce is reviewed by a human before it ships.

Corrections Policy

NUZU does not write articles, so there is nothing for us to correct in the traditional sense. But we do make decisions about what appears on the site, and those are worth getting right.

If you believe something on NUZU is wrong — a miscategorized source, a broken link, a tier that no longer reflects a publisher's standards, an article linked in error, or a publisher that wants to be removed — contact us and we'll fix it.

1. How to reach us: NUZU-NEWS@protonmail.com.
2. What to include: the URL of the page, a screenshot or link to the item, and what you believe is incorrect.
3. What we do: we review within 48 hours. If the issue is with something the site itself did (wrong tier, wrong section, a duplicate, a broken link), we fix it on the next build and note the change in this document's revision log.
4. What we don't do: we don't correct, retract, or argue with the content of articles published by third-party outlets. Those corrections belong to the publisher that wrote the article.
5. Publisher removal requests: any publisher that prefers not to be linked from NUZU can email us and we will remove their feed on the next build.

Your data stays yours

We don't run analytics. We don't set tracking cookies. We don't require an account, login, or email address. Your saved articles, theme, font size, and section preferences are stored only in your own browser — never on any server. See our Privacy Policy for the details.

Install it as an app

NUZU works as a Progressive Web App. On Chrome, Edge, or Android, tap the install button that appears after your first few visits. On iPhone, use Safari's "Add to Home Screen." Once installed, NUZU launches in its own window and works offline for headlines you've already loaded.

← Back to the headlines