Product principles
Two Doors: the dark pattern we refused to build
A small moment of chance became a useful test: would we make a promise the product could not honestly keep?
Two Doors occasionally offers two unlabeled doors before the next page. You choose one without knowing what is behind it. The pages are simply the next candidates already queued up; they are not reliably different in depth, tone, or surprise.
The version we did not ship
It would have been easy to label the doors: “Quick read” and “Deep dive,” or “Something familiar” and “Something surprising.” That would make the choice feel more meaningful. But the app cannot know those things about the two candidates at that point. A label would be a promise without evidence.
So the doors are just Door A and Door B. It is a small ritual of chance, presented honestly as chance.
What makes this a dark-pattern question?
A dark pattern does not have to be visually aggressive. It can be a label that makes a choice feel informed when the product has no evidence behind it. In this case, calling one door a deep dive or a familiar read would have turned an ordinary choice between queued pages into an implied prediction.
Honest product design does not mean removing surprise. It means making the boundary clear between what the product knows and what it does not.
Why that matters
There is usually a version of a product decision that improves engagement by making a user feel more certain, more compelled, or less able to leave. Serendipity is designed around declining those versions when they ask the user to trust a claim we cannot support.
That is why it suggests a stopping point after a good run instead of inventing a reason to keep browsing. It is why a topic search says it does not have a confident result rather than lowering the bar. And it is why its delete-everything button does what it says.
The product underneath the feature
The goal is not to make browsing sterile. It is to make discovery enjoyable without turning uncertainty into a trick or attention into the business model.