Reading Claims

A Claim is one research conclusion, written down once in a structured way and attached to the things that back it up. A paper says "here are forty pages of work". A Claim says "here is the one thing we now know, and here is exactly why you should believe it".

Nobody hand-picks Claims. PIPress continuously reads public sources (code repositories, research archives, model hubs, academic sources, open discussion) and runs each item through a sequence of automated checks. Each check can reject what the one before it passed along. Only what survives becomes a Claim, gets scored, and reaches the feed.

The score and the tier

Every Claim carries a score, shown as a percentage, and a tier letter from A to D:

Tier	Shown as	Score
A	Verified	79% and above
B	Corroborated	53% to 79%
C	Contested	26% to 53%
D	Speculation	below 26%

The score answers one question: how much does this actually change things? It is not a quality rating or a popularity count. A ratified standard that is already deployed scores near the top. An interesting idea in a draft scores low, not because it is bad, but because nothing has structurally changed yet. The biggest ingredients are how final the change is, how good the evidence is, and how much it connects to other fields.

Separate from the score, each Claim has a confidence of High, Medium or Low: how settled the underlying finding is. A Claim can matter a lot and still be early.

Anatomy of a Claim page

Open any Claim (click its title in the feed, or share its link, which looks like pipress.xyz/feed/<id>) and you get the same sections every time:

1. What happened. A factual reconstruction of the event.
2. Analysis. Why it matters structurally.
3. Counter view. The strongest opposing argument, deliberately included.
4. What we don't know. What remains unresolved.
5. Sources. Every piece of evidence, labelled by how direct it is: a primary source (the specification, the peer-reviewed paper, the merged code) counts for more than a narrative one (a blog post or a talk). Only links that really appear in the source material are kept.
6. What this assumes.
7. How we reached this. The reasoning, step by step.
8. Related topics and How this connects across layers. Typed links to other Claims: one Claim builds on, contradicts, depends on or replaces another. Every proposed link is checked before it is kept.
9. Where this came from. The raw items that were detected and used, with their source types.
10. Revision history. Claims are living documents. When new information arrives, the Claim is updated rather than duplicated, and every version is kept.
11. Discussion. Public comments on the open Nostr network.
12. More in this domain.

The panel on the right shows the score, the tier, when the Claim was first seen and published, its domain, its confidence, the source types and the key entities.

What you can do with a Claim

• Share. Copy the link, or send it to X, WhatsApp, Telegram or a Nostr viewer.
• Flag. Opens an email to the team if something looks wrong.
• Hide (signed in). The Claim disappears from every list for you, with six seconds to undo. Hidden Claims can be brought back under Settings → Hidden claims.
• Useful / Not useful and Zap appear on Claims that have been published to Nostr. These are signed with your Nostr key, so an email-only account is asked to link a key first. Zaps are small Lightning payments sent to the platform's address from a wallet you connect or by scanning a QR code. See Settings.

Everything you do here (open, copy, share, mark useful, zap, hide) teaches your personal feed what you care about. See Your feed.

Browsing by subject

• Ecosystems lists every domain PIPress covers across the two tracks, with its topics, a twelve-week activity chart and the number of active Claims. Click Enter to see a domain's top Claims by score.
• The Knowledge Map draws the domains and the topics inside them as a graph built from published Claims. Select a domain to load its topics. Below the map you see how confident the Claims are overall and whose work they come from.