Why we don't use an algorithm.

The default news app in 2026 hides chronology behind a model. Headlines aren't shown in the order they were published; they're shown in the order an engagement model thinks will keep you in the app. The feed is one of the most consequential design decisions on the public internet, and most users have never seen what it would look like without the model.

WeSearch shows you that. The home feed is chronological, deduplicated by URL, and identical for every reader. This page lays out why we made that choice and what you give up to take it.

What an "algorithmic" feed actually is

An algorithmic news feed is a ranked list. You arrive at a page; the model takes a candidate set of recent stories, scores each one with a function that approximates "this user will engage with this", and returns the top-N in score order. The function is trained on prior user behavior: clicks, dwell time, scroll depth, share rate, return visits.

This sounds neutral. It is not. The training signal is engagement, not informedness. A story that produces an angry click but adds nothing to your understanding of the world will outscore a slow, careful piece that takes ten minutes to read. Over a year, you will have read more of the first kind than the second.

The four problems with engagement ranking

1. The bubble. Personalization narrows. The model learns what you click and shows you more of that. The set of stories you'd also have engaged with — but didn't because you never saw them — shrinks. After a few months, the difference between your feed and a similarly-aligned reader's feed is invisible to you because you both stopped seeing the other side.

2. The optimization mismatch. "What will make me click" and "what is news" diverge. A model that maximizes the first is, by definition, not optimizing the second. You can paper over this with editorial intervention, but at scale, the model wins.

3. The opacity. Once the order is opaque, the publisher loses agency. A newsroom can't predict whether a piece will reach its audience because the audience is downstream of a model the publisher doesn't see. This bends reporting toward whatever traits the model rewards.

4. The asymmetric power. Whoever runs the model picks the news. If the model is trained on engagement, the engagement-maximizers pick. If it's run by a platform, the platform picks. The reader, who actually has skin in the game, is the only party with no input.

What chronology gets you instead

Chronology is dumb. That is its strength. A chronological feed has no opinion about which story should be on top — the wire's publish time has it. Every reader sees the same feed, so the platform can't tilt one demographic's view of the world without tilting everyone's. The publisher gets exactly the discoverability the publish time earns. There is no model to game.

What you give up is convenience. A chronological feed will sometimes show you headlines you don't care about. That is the point. A serious news reader needs occasional encounters with stories outside their priors, because the alternative is a feed that has slowly lensed itself onto a single worldview. The dumb sort buys you the encounter.

What about "trending"?

WeSearch has a trending row, but it is a count, not a model. The 24-hour trending list ranks stories by the number of distinct anonymous handles that reacted to them. There is no velocity boost, no virality smoothing, no engagement-prediction layer. If a thousand people react to a story, it surfaces. If two people react to a story, it doesn't. That is a count, not an algorithm.

What about category filters?

You filter the feed yourself. 700+ sources across world, politics, tech, science, business, markets, health, climate, ai, culture, sports, and so on. The categories come from the source's own classification, not from semantic clustering. You explicitly pick what you want; you can see exactly which sources contributed; you can opt in or out at the source level.

What about discovery?

The community layer handles discovery. The Pulse tab surfaces what other anonymous readers are reacting to. The "voices in the room" panel surfaces the most-engaged commenters this week. The Most Discussed row surfaces stories that drew real conversation, not just clicks. The signal is community engagement, transparently counted, not a model output.

The honest trade

You will, with chronology, sometimes scroll past stories you don't care about. You will, occasionally, see headlines that other readers in your political tribe have decided to ignore. Your feed will not be optimized to keep you in the app. Those are the costs.

The benefit is that what you read is what is happening, in the order it happened, across many sources at once, without a third party deciding which version of "the news" you get. That trade made sense for fifty years of newspapers and we think it still does.

Frequently asked