Why We Built B.I.A.S.

The news has always reflected the world. What has changed is which parts of the world it chooses to reflect.

Pick up a newspaper from thirty years ago and you will find stories about elections and disasters sitting alongside features on unusual trades, wildlife surveys, and the opening of a new bridge in a town most readers had never visited. The ratio felt different. There was room for the world to be interesting, not just urgent.

Somewhere along the way that changed. The web did not cause it — economic incentives did. Clicks reward anxiety. Shares reward outrage. Algorithms favour the story that makes you feel something sharp and immediate, and the easiest emotion to provoke is fear. The result is a news environment that selects, systematically, for the most alarming interpretation of every event. Most of us know this. Most of us keep scrolling anyway.

B.I.A.S. is an attempt to select differently.

We built an automated pipeline that scans around fifty international RSS feeds and dynamic search topics several times a day. It scores every candidate story against a rubric we have refined over many months:

The highest-scoring story that clears a paywall check gets selected, summarised in the site's editorial voice, and published. One story, roughly every two hours.

What we reject

The algorithm is as much about what it excludes as what it finds. Stories about political conflict, even when reported calmly, tend to score poorly — not because politics does not matter, but because there are already hundreds of outlets covering every development in real time. We do not need to add to that noise.

We also reject stories where urgency is the main product. Breaking news, developing situations, hourly updates — these are designed to keep you anxious and checking back. We are not interested in that transaction.

Sensationalism scores poorly by design. A headline that exists to make you feel fear, contempt, or tribal satisfaction is doing something to you that we would rather not participate in.

What we are looking for

The stories that score well on B.I.A.S. tend to share something: they feel like they could have been published last week, or six months ago, or possibly two years from now, without losing much. They are not urgent. They are interesting. They reward the five minutes you spend with them.

A restoration project in a Japanese mountain village. A study on how certain dialects preserve grammatical structures that died out elsewhere centuries ago. A marine biologist who spent a decade mapping a coral reef that most people have never heard of. A craftsperson who is probably the last person alive who knows how to do a particular thing.

These stories exist. They are published every day, in outlets across the world. They are just not the ones the algorithm usually surfaces.

How it works

B.I.A.S. is a Python pipeline running on a home server. It fetches feeds from publications in Europe, Asia, Australasia, and the Americas — including sources in French, German, Spanish, Japanese, Arabic, and Chinese, with automatic translation where needed. Each candidate is scored by a language model trained to recognise the editorial qualities we care about. The top scorer gets its article text extracted, checked for paywalls, and summarised. Then it is published.

There is no editorial board. No advertising pressure on the selection process. No engagement metric to optimise against. Just a rubric, a model, and a preference for stories that do not make you feel worse about being alive.

You can read more about the sources we draw from on the Sources page, or get in touch via the contact page if you have a publication to suggest.

The name

B.I.A.S. — Balanced Information, Actual Stories — is a deliberate irony. We are biased. Every editorial process is. We just try to be biased toward things worth reading. Read about where we want to take that idea next on the What's Next page.