About

What this is

Alivewire is a reading desk for health and longevity news. Every twenty minutes it checks several hundred publications, journals, institutions, newsletters and podcasts, groups the pieces that report the same story, and links each to the outlet that wrote it. Nothing is republished; every headline on the site leads out.

What it is not

Alivewire is not medical advice and does not attempt to be. It records what is being written and by whom. Whether a finding bears on any one reader is a matter for that reader and a clinician, and a single study, however widely covered, rarely settles a question. The evidence chips described below exist to keep that in view.

Where the stories come from

Sources are catalogued by hand and tiered. Tier A outlets (the specialist health press, major newspaper health desks, the big journals' news feeds, a handful of longevity-specific publications and podcasts) are on by default. Tier B is on by default as well. Tier C, which includes consumer wellness brands, trade press and high-volume general feeds, is fetched but off by default, and can be enabled in the Sources panel. High-volume outlets are capped per day so no single feed floods the stream.

Each source carries a reach label: mainstream (national newspapers, broadcasters, wire services), specialist (health and science press, journals, institutions) or independent (newsletters, podcasts, individual researchers). Stories from outside the US say which country they come from, and the Region filter narrows the whole page, trending names and sources list included, to North American or European outlets.

Journals and preprint servers are filtered to aging and health topics before they enter the stream. Non-English stories are shown with an English headline and snippet; translations are added by hand during a daily editorial pass, and until a story has had its turn it carries a translation pending chip. The Language button in the header switches between the translation and the original.

How stories are grouped and ranked

Pieces about the same event are grouped into one story, and the story page lists every outlet in the order they reported it. The stream ranks stories by how many outlets are covering them, decaying over about a day and a half, with a floor so that small topics are not buried by big ones. Default-off sources count for less, and so do very high-volume outlets: a feed posting sixty pieces a day cannot outweigh five specialist desks by volume alone. There is no engagement signal of any kind in the ranking; what readers click plays no part in it. The site uses Google Analytics for aggregate traffic measurement.

Compounds, researchers, institutions, trends, products

The names in the sidebar come from a curated vocabulary of around 1,300 compounds, interventions, researchers, institutions, companies and trends, plus a small set of patterns for new names (a doctor or professor named in a headline, a university). New names are reviewed during the daily pass and wrong ones are removed. Each name has its own page with its coverage over time and the names that appear alongside it. Products is a separate channel, off by default, for readers who want to follow what is being launched or reviewed; product-only stories enter the main stream only when the channel is enabled.

Evidence chips

Study-shaped stories carry small chips describing the kind of evidence behind them, read off the headline and snippet:

• meta-analysis pools several studies. randomized trial assigned participants at random, the strongest design for cause and effect. observational means an association was found, not a cause.
• peer-reviewed names the journal when the story does. preprint means the paper has not been peer reviewed yet and can change.
• in animals and in cells flag work that has not been done in people. n=1,200 is the number of participants when the text gives one.

The chips are deliberately literal and only as reliable as the headline and snippet they are read from. A story that does not say how a study was conducted receives no chip, which is itself a small signal. Errors are corrected by hand.

Topic explainers, evidence stage and signals

The most-covered compounds, trends and products carry a short explainer at the top of their page: what the thing is, why it is in the news, where the evidence stands. Explainers are written by the editors from the coverage, with reference sources such as regulator labels and Wikipedia for background, and are dated. They are descriptions of the record, not recommendations.

Compounds, products and intervention-style trends also carry an evidence stage, a 1 to 10 score for how settled the human evidence is for the topic's main use. The number is a rung on this ladder, not a measurement, and a note under the explainer says which use it describes and where the same thing is still experimental (a drug can be a 10 for diabetes and a 4 for alcohol use disorder):

• 1–2 preclinical: mechanism, cell and animal work only; no meaningful human data.
• 3–4 early human evidence: small, short or observational human studies; no randomised trials of note.
• 5–6 mixed or limited trials: one or a few randomised trials, mixed results, or trials in narrow groups.
• 7–8 established for some uses: several randomised trials or approval for at least one use; open questions remain.
• 9–10 thoroughly established: decades of use, meta-analyses, a place in clinical guidelines.

Where one number would mislead, the page adds evidence by use: a short breakdown that scores each major use or claim on the same scale. Stem cell therapy is the model case. Blood stem cell transplants for leukaemia have been standard care for decades, while injections sold for joints, ageing and autism have little or no trial support, and the breakdown shows that spread rather than hiding it inside an average. An "Uneven" chip beside the headline score marks these topics.

Signals are deliberately not a safety score. We do not rate how safe anything is, because that depends on dose, duration, the person taking it and what else they take. Instead the explainer lists discrete facts from the record, each tagged by kind and linked to a source where there is one: regulator actions (boxed warnings, alerts, recalls, counterfeit warnings), trial outcomes (negative, neutral or halted trials, adverse effects seen in people), mechanism questions (concerns argued from how something works, not yet seen in people), and market status (approved, off-label, compounded, grey market, unregulated). The chip summarises the strongest kind present: no signals reported, theoretical, observed in trials, or regulator warning. "No signals reported" means nothing in what we read, not proof of safety.

Ratings are revisited during the daily pass when a large trial, approval or warning lands, and the date of the last rating is shown on the page. Readers who believe a rating is wrong can reach the editors through the contact on this page.

Editorial pass

Each day an editor reviews what has come in: merging duplicate names, muting misbehaving feeds, translating headlines, correcting groupings, and removing sources that prove to be marketing. The site runs continuously; the editorial pass follows behind it. Featured stories are labelled, and if a placement were ever sponsored it would say so on the card.

Feeds and editions

Every view has an RSS feed: the front page, each topic, each compound or researcher page, and any filter combination (/feed.xml?kind=study&sections=aging). A daily edition fixes the previous 24 hours at 7am Eastern as a permanent page. There is a JSON API behind the site as well; the By publication page lists everything read.

Alivewire is an independent editorial project. The feeds, the vocabulary and the grouping are all open to correction; the editorial pass is where it happens.