RS / Signal

v3.7.0Search

Signal is a public news brief. It reads 98 hand-picked publisher feeds, groups the same story told by different outlets into one record, and files 12 records every hour. It has no feed, no recommendations, no engagement metrics, and no advertising.

The crawler

Every request Signal makes carries this User-Agent, which is how you identify it in a log:

SignalBot/1.0 (+https://signal.sharmaritik.com/about)

Requests: One per feed per run, 96 runs a day
Conditional: If-None-Match and If-Modified-Since, whenever you gave us one
Response cap: 8 MB, then the connection is dropped
Robots: Only feeds you publish; no crawling of article pages

Conditional requests are the point rather than an optimisation. Each validator is sent back only if your feed supplied it, so a first fetch and a feed that returns neither header are full requests; every fetch after that carries whatever you gave us. A feed that has not changed then answers 304 with no body, so the overwhelming majority of those runs cost you a header exchange and nothing else.

What is stored

Kept: Headline, lede, link, publication time, publisher name
Kept: A vector computed from headline and lede, for grouping
Deleted: Article body text, after 72 hours
Never stored: Images, video, full article text beyond that window

Body text is a processing artefact. It is never rendered to a reader, never rendered to the operator, and is deleted on a schedule rather than on request. What a reader sees is a headline and a link to you.

What is published

Records: 12 per filing, never more
Filings: 24× daily, on the hour, Australia/Brisbane
Every record: Lists every publisher behind it, each linked
No: Reader mode, republished prose, or content behind our chrome

Some records carry a short synthesis — WHAT HAPPENED, WHY IT MATTERS, WHAT CHANGED— written by a language model from the headlines of the sources listed on that record, and labelled as machine-written wherever it appears. It is assembled from what those sources state rather than reproduced from any one of them, and no publisher’s article text is republished. Records in high-liability categories — deaths, criminal allegations, active legal proceedings, medical claims about named individuals — never carry one, by rule rather than by judgement. Where a record has more than one source, the additional sources are evidence that the story was corroborated, not material to reproduce.

Removal and contact

To have a feed removed from the registry, or to ask anything about the traffic, use the contact page. Removal is a registry edit and takes effect on the next run — there is nothing to negotiate and no form to fill in.

← THE BRIEF