Interface Crimes · No. 001
Search as an evidence board
Search has been a box for thirty years. This entry treats results as evidence instead of answers — pinned to a wall, connected by relationship, allowed to disagree with each other and with you.
Intent
Search starts with a guess, then asks what kind of truth you need.
Capturing user intent as a first-class signal.
Persistent context
Results inherit context from what you inspected yesterday.
Why context should follow you across tools and time.
Contradiction
The most useful result may be the one that disagrees with you.
Search becomes less obedient — and maybe more honest.
Synthesis
Ranking becomes a visible reasoning structure.
From finding answers to building understanding.
Four sources, two camps. The red thread marks where they disagree.
What it breaks. One-dimensional ranking, passive scanning, and the assumption that relevance is linear. A ranked list makes research look like a queue; understanding is built through comparison.
Why it might work. Research tasks need contradiction, memory, and spatial grouping. As AI agents do more of the searching, the interface may need to show reasoning, provenance, and source tension rather than only results.
Where it fails. Fast lookup. Accessibility-heavy workflows. Any case where the answer is obvious and a box would have done fine.
Would this make research clearer — or just turn every search into a detective wall?