How we work
A model drafts. Code checks every number. A person approves every article.
How it works
- ReadCode reads Hyperliquid’s public data. No model guesses a number.
- DraftA model writes the text from those numbers, in English.
- CheckCode matches every number in the text to the data, and the words to our rules.
- ApproveA person reads it and approves it. Nothing goes live without that.
Our rules
- Facts and numbers, with their source and time in UTC.
- No forecasts, no advice, no promises of returns.
- Other outlets’ stories are not reprinted: a fact, a number, a link to the first source.
- Corrections stand on the article, dated, at the top and the bottom.
What the model does and doesn’t do
The model writes sentences around numbers that code has already read. It has no say in what a number is: if the text holds a number that is not in the article’s data, the article stops before a person sees it.
A person reads every article before it goes live and can change or drop it. Nothing is published automatically.
Mistakes
If we get something wrong, we fix it and say so on the article: what changed and when. The date of the change shows beside the date of publication.
