How it learns your taste

Lookout watches what you do:

Action	Teaches
Star, done	strongly: more of this
Open (first time)	mildly: more of this
Dismiss	less of this
Snooze	nothing

It learns weights for sources, authors, categories (and the ones above them) and tags. Older actions count less (half-life of 30 days by default), and nothing is learnt until you have acted at least five times.

What it changes

Items are ordered by priority first, then by model score plus a bounded adjustment of at most 20 points from your profile. The adjustment only orders items inside a priority: a P0 stays P0 even if you dismissed everything from its source.

Why this rank

Open an item in the dashboard, or run lookout explain <id>:

P1 (this week): model score 85: You write Python.
model score 85
your profile moves it +16 inside its priority
  +16 category:Dev > Rust

Starting over

lookout forget (or DELETE /api/profile) resets what was learnt. Items and their states stay.