Editorial standards

Voice

variis writes in one voice. It is confident, internationally fluent, and modelled on Bloomberg, Monocle, and FT Weekend. Sentences are declarative. Claims are specific. Restraint is a brand value, not a stylistic accident.

We do not use the language of hype. There are no breakthroughs unless something has actually broken through, no revolutions in supplement aisles, no miracle molecules. If a finding is preliminary, we say so in the same sentence that describes it. If a company is early, we describe it as early rather than as a category leader. Editorial restraint is the written form of our monochrome design system: the story carries the emphasis, not the adjectives around it.

The author of every article is the institution. The byline reads variis. The reviewer, however, is named. That asymmetry is deliberate. It signals that the reporting is the work of a publication with a house standard, while the medical judgement behind publication belongs to an accountable individual.

Sourcing

Every factual claim in a variis article traces to a named, dated source. Sources are listed at the foot of each article in the order they are cited, with publication date, publisher, and a direct link. Where a source sits behind a paywall or on an ephemeral news page, we include an archived snapshot URL alongside the original.

We do not write "studies show". We name the study, the journal, the year, and the sample size. We do not attribute claims to anonymous experts unless there is a clear public-interest reason and the identity is known to the editorial desk. Preprints are labelled as preprints in the body of the article, not only in the footnote. Regulatory filings, clinical trial registrations, and company disclosures are cited by their registry identifiers where they exist.

When a source is a company itself (a press release, an investor deck, a founder interview) we say so in the sentence that cites it. Marketing material is treated as marketing material, whatever the market cap of the marketer.

The four-gate credibility model

Every variis article passes through four gates before publication. Three are handled by Claude models under editorial supervision. The fourth ends with a named human. A story that fails any gate does not advance until the failure is resolved.

Gate 1: Triage

The first question is whether the signal is a real, novel Singapore or Australia longevity story, or a duplicate, stale, or off-topic input. Triage filters press releases we have already covered, generic wellness content, and stories whose geographic or thematic relevance does not clear our bar. This gate is handled by Claude Sonnet 5, with human oversight on ambiguous cases where the news value is uncertain.

Gate 2: Draft

Once a story is cleared for coverage, a draft is generated from the supplied sources and only from those sources. Every claim in the draft is tied to a specific source URL as it is written, so that the provenance of each sentence is visible to the next gate. Drafting is handled by Claude Sonnet 5. The model is instructed to leave a claim out rather than to reach for a fact that is not in the source pack.

Gate 3: Verify

Verification is a claim-by-claim audit. For each factual statement in the draft, the cited source is opened and checked to confirm that it actually supports what the sentence says. Claims that the source does not support are removed or reworded to match what the source actually establishes. Numbers, dates, names, affiliations, and study designs are all re-read against the original. This gate is handled by Claude Opus 4.8, chosen for its higher reliability on close reading of long documents.

Gate 4: Brand and compliance

The final gate is a two-part check. Automated scans confirm that the draft does not contain banned voice terms, that the brand name variis appears in lowercase throughout, that em-dashes have not been introduced, that any medical or dietary content carries the required YMYL disclaimer, and that the evidence grade attached to the piece matches the strength of the sources cited. The draft is then reviewed by [Reviewer Name] before publication. Nothing publishes until that human review is signed off.

Evidence grading

Every variis finding carries a four-bar evidence level, shown on the page next to the claim it grades. The scale is deliberately conservative.

• Mechanism (1/4). A plausible biological mechanism has been proposed. There is no direct evidence in humans, and often none in animals. A hypothesis worth watching, not a result to act on.
• Preclinical (2/4). Evidence exists in animal models, cell lines, or other in vitro systems. Human trials have not yet begun or have not yet reported.
• Clinical (3/4). Human clinical trial evidence exists. We describe the trial phase, sample size, endpoints, and whether results are peer reviewed.
• Commercial (4/4). The intervention is approved by a relevant regulator, or is otherwise in ordinary commercial use, in at least one jurisdiction we cover.

The grade is a check on our own excitement. A large funding round is not a scientific result. A published paper is not a clinical outcome. A regulatory clearance in one market is not a global standard of care. The evidence meter says so on the page, in a form the reader can see at a glance.

Named review

Every article that publishes on variis has been read and approved by [Reviewer Name]. The byline is institutional. The review is named. This is our answer to the requirement, in any responsible publication that touches health, that a credentialed human is accountable for what appears under the masthead. When the reviewer is uncertain, the article does not publish. When the reviewer is on leave, publication pauses.

Corrections and disclosures

Corrections are dated and published publicly at /corrections. When we change a substantive fact in an article after publication, we say what was changed, when, and why, and we link the correction from the article itself. Silent edits are forbidden. Typographical fixes are logged internally but not surfaced.

Disclosures are shown inline. If variis, its editorial staff, or the named reviewer has any commercial relationship with a company, institution, or product discussed in an article, that relationship is disclosed at the top of the piece in plain language.

Contact

Questions, corrections, and tips reach the editorial desk at editorial@variis.co.