The five rules
An article does not go live if it misses a single one.
| Rule | How it is checked |
|---|---|
| 1Every factual claim has a checked source | The URL was opened and says what it is made to say. No reference cited from memory. |
| 2The article challenges at least one received idea, and explains why | If a reader who knows the subject learns nothing, the article is filler. |
| 3It contains at least one element nobody else has | A table, a protocol, a documented limit, a figure from our database. |
| 4It states explicitly what it does not say | A "what this article does not say" section closes every article. |
| 5It is dated, signed, and updated when the literature moves | Publication and update dates visible, identified author, public corrections log. |
What counts as a source here
A source is a primary document: the trial, the meta-analysis, the standard, the agency opinion, the regulation. Not the press article about it, not a vendor's page, not a second-hand summary. Every source is opened before it is cited, and the sentence in the article links from the exact word it supports, clickable inside the paragraph.
When a source has a history, the article tells it: a trial withdrawn then republished after randomisation irregularities, a result that is only a secondary outcome, two trials sharing an author and a branded product. An invented or approximate source is worse than no source at all.
What a source says and what it is made to say are two different things. A study measuring a wearable's agreement with polysomnography on a given night says nothing about that wearable's ability to track a change within the same person. Making a source say what it did not measure is the least noticeable violation of rule 1, and the one we watch most closely.
Where personal experience fits
The founder tests his own protocols and writes about it. A testimony is not a source, and rule 1 allows no exception, but a testimony is not nothing either. It is announced for what it is from the first lines, it is not retracted in the next paragraph, and it is confronted with the published result that troubles it most, without that result being taken out of context.
Not all personal observations are equal, and every article says which level its own sits at:
| Level | What it is | What it is worth |
|---|---|---|
| Impression | "I feel better" | A starting point, nothing more. |
| Repeated observation | several with-and-without alternations, stable conditions, same measure on the same device | Below a controlled trial, above an impression. It gets said, it does not get swept away. |
| Convergence | an unexpected observation plus a documented mechanism that explains it | Proves nothing, but makes the hypothesis serious. |
A limit set against a testimony must say what it invalidates and what it leaves standing. When a testimony survives its limits, it is left standing: the article's job is to place it, not to win against it.
Corrections
Every correction received is logged, the rule it reveals is added to this method, and when it can be detected, an automated check verifies it across all articles before publication. A correction on today's article applies to all the previous ones: it is applied retroactively.
Changes that alter what an article claims are listed publicly, with date and reason, on the Corrections page. An error reported by a reader appears there the same way. To report one: contact.
What this blog is not
- Medical advice. Helix is an educational tool, not a medical device, and nothing here replaces a health professional who knows you.
- An advertising medium. No ads, no affiliate links, no sponsored content. When a branded product is named, it is because a trial used it, and the article says so.
- Content reviewed by a doctor. It is not, and we do not claim it is. The day a health professional reviews an article, their name will appear on it.
Who writes
All articles are written and signed by Théo Sans, founder of Helix Health, who builds the project alone and in public. His author page gathers his background and all his writing.