How we decide what to publish

What we publish

HafidWatch covers news that affects crypto markets, macro conditions and global finance. Each article is a short analysis of one published report: what was reported, who said it, what the figures show when set against each other, and what remains unknown.

HafidWatch does no reporting of its own. We do not interview sources, obtain documents or break news. Our contribution is selection, context and clear explanation.

How stories are selected

An automated system reads our source list every hour. Each new story receives a score from a fixed formula based on the outlet, the topic, how recent the story is and the headline. Sponsored posts, promotions, weekly digests, opinion columns and price-prediction pieces are filtered out. The highest-scoring story that we have not already covered moves on to the writing step.

How articles are written

Articles are written by an AI model working from the text of a single published report. The model follows written rules that we maintain.

Every fact, figure, date and quotation must appear in the original report. Nothing is added from memory.

Claims are attributed to whoever made them: the company, the regulator, the filing, the data provider or the person quoted. Allegations are reported as allegations.

When we calculate something from the reported figures, such as a ratio or a difference, the sentence says so and shows the inputs.

Our own interpretation appears in a separate box labelled “HafidWatch Take”. It is built only from the reported facts and is never a forecast.

The outlet whose report we worked from is named in a source line at the end of every article. Price data shown alongside an article comes from CoinGecko and carries a timestamp.

Automated checks before publication

A draft is published only if it passes every check. The original report must contain enough material to support an article. The draft must reach a minimum length without padding. The model must also grade its own draft, and anything it marks as needing revision or as not worth publishing is held back. Drafts that fail are discarded, and the story is not retried.

These checks reduce errors. They do not eliminate them, which is why every article is also reviewed by a person.

Human oversight

The source list, the scoring formula, the writing rules and the checks are all written and maintained by a person. Individual articles are published automatically once they pass the checks, and are then reviewed by that same person.

Sources we monitor

We currently monitor seven outlets, in two groups. This is the same grouping we use for the Divergence Index. The list changes over time, and this page is updated when it does.

What we never publish

• Price predictions or price targets of our own, or advice to buy, sell or hold
• Facts, figures or quotations that are not in the report we worked from
• Promotions, giveaways or token marketing presented as news
• Clickbait headlines, or claims stronger than the evidence supports
• Statements that HafidWatch confirmed, verified or obtained something it did not

A forecast made by a named person or firm may be reported as theirs, with attribution.

Sponsored content and affiliates

All sponsored content is clearly labeled “Sponsored.” Advertisers and sponsors have no say in which stories we cover or how we cover them. We have affiliate relationships with Ledger, TradingView and NordVPN and may earn a commission from links to them. Affiliate links are disclosed where they appear.

Corrections

When we make a factual error, we correct it promptly. Corrections are noted within the article with the date of the correction. Articles that do not meet our standards may be removed. See our full Corrections Policy →

For how we use AI in detail, see our AI & Automation Disclosure →