The billing model

What runs: the reading meter

Every page carries <sg-meter>, SG Meter v1.3.1, vendored under assets/components/ and loaded from this site only. You start with £5.00 of credit in this browser; the balance is in the top bar and goes below zero without anything being blocked. A page costs its price times the share of it you scrolled through, rounded to 5% and never less than a tenth, once per page per session. At the foot of each page you can rate how useful it was, which sets the price from free to double, or decline to pay with a reason. Nothing leaves the browser: no account, no cookie, no server of ours, no analytics listening to the meter's events. Your reading account shows the history and exports it as JSON.

The formula

price = max(floor, round(base[type] x freshness[tier])), then x the share of the page read (never under a tenth) x the reader's usefulness rating, with a floor of 1p.

A dot means the price does not decay: the type costs the same at any age. The date the page's content last changed: for a rebuilt page, the time of the latest capture its data comes from (so new evidence makes a page fresh again); for an article, the date it was written (so an article only ever ages).

The rules of engagement

1. Every content page is metered. The only unmetered pages are the meter's own: your account, top up and topped up.
2. The price depends on two things only: what kind of content the page is, and how fresh it is. Not on who is reading, not on demand, not on the topic.
3. Fresh costs more and the price halves into the archive, never to zero. Evidence and the site's account of itself do not decay and stay at the floor.
4. A page costs its price times the share of it you read, never less than a tenth; once per page per session; rating it re-prices it from free to double.
5. Nothing is ever blocked. The balance can go below zero, and a reader can decline to pay and say why.
6. The age is worked out in your browser from the date on the page; the build has no clock, so the same release serves the same prices table to everyone.
7. The prices table is published here and generated from the same data file as the meter, so the page and the meter cannot disagree.

Fresh. The first day: when the content is news and a reader is deciding something now.

New. The first week: still the current state of a developing story.

Current. The base price: settled enough to rely on, recent enough to act on.

Archive. The record: worth reading, priced for the reader doing history, never free.

Top-ups are £5.00 of credit on a payment page. No payment link is configured for this site yet, so the top-up page runs the meter's simulated cart and says so. When a link is added, the credit is added on the return page without verifying the payment, because a static site has nowhere to receive a webhook; the meter's security model is explicit that this is an honesty box, not a paywall.

What is designed: paying for provenance

• A price per verified claim. Each claim on this site carries a status and the sources that give it that status. The unit a reader or an agent would pay for is not the page but the claim with its evidence chain: 'is this use of it sound', in the newsroom's phrase, not 'is this true'. Proposed price: 1p per claim read, 5p per claim with its captures opened, 20p per claim re-verified on request.
• The fact creator's share. The newsroom's economics page argues a 60/25/10/5 split between the creator of the fact, the checker, the publisher and the rails. On this site the creators are the agencies, the researchers and the reporters whose pages are captured; the checker is the collector agent and the verifier; the publisher is this site. The split is argued, not implemented: nothing here pays anyone.
• Reading the vault itself. A reader who wants the bytes rather than this site's reading of them would be sold a read-only snapshot of the evidence vault for a fixed sum, opened in the browser with a read key. The captures are third-party pages and are not for republication, so this stays a proposal until the rights question has an answer.
• An agent's subscription. An agent that briefs an executive every morning would pay per briefing regenerated, metered by the same component, with the balance kept in the agent's own vault rather than a browser. The meter already fires an event per charge; nothing listens to it here.

None of this is live. The newsroom's pages on paying the fact creator and provenance as the product carry the argument; this site is the first with a claim table granular enough to price.

What this story cost to assemble, as a model

A model of what this story cost to assemble, in units counted from the vault and the repository. The unit prices are assumptions, labelled as such; the counts are not.

Total at those assumptions: £130.65. The counts are from the data files named; the prices are the site's assumptions and a reader is invited to substitute their own. What the model leaves out is the editor's time and the model inference behind each agent run, neither of which is counted anywhere yet; the honest column on this is not measured.

The prices, as data

{
 "prices": {
  "account": 0,
  "analysis": 3,
  "analysis_archive": 2,
  "analysis_current": 3,
  "analysis_fresh": 6,
  "analysis_new": 5,
  "article": 6,
  "article_archive": 3,
  "article_current": 6,
  "article_fresh": 12,
  "article_new": 9,
  "briefing": 5,
  "briefing_archive": 3,
  "briefing_current": 5,
  "briefing_fresh": 10,
  "briefing_new": 8,
  "evidence": 1,
  "site": 1,
  "story": 2,
  "story_archive": 1,
  "story_current": 2,
  "story_fresh": 4,
  "story_new": 3
 },
 "start": 500,
 "storageKey": "php-story.meter.v1",
 "symbol": "\u00a3",
 "topup": {
  "amount": 500,
  "link": ""
 }
}

This block is data/billing.json, which is also what every page's meter config is generated from, so the table above and the meter cannot disagree.