Slot

The attribute a claim fills — employer, diet, commute — as opposed to the words it uses. Two beliefs compete when they fill the same slot for the same subject.

Why it matters in a memory layer

Slots are how conflict candidates are found, and the reason is measured rather than aesthetic: in this course's corpus the employer contradiction scores 0.285 on cosine similarity, below unrelated noise at 0.478 and far below a refinement at 0.669. Two claims can disagree completely while sharing almost no wording — "data engineer at Northwind" and "works at Calico Systems" overlap in one word. Any similarity threshold that catches that pair catches everything. Grouping by slot finds it directly, and shrinks the candidate set at the same time.

Connections

Taught in: Contradiction Detection

Used in: Component Metrics · From Facts to a User Model · Provenance and Trust · Reflection and Insight · Scaling the Store · Sleep-Time Compute · What Must Never Be Dropped · Hybrid Ranking · ADD, UPDATE, MERGE, NOOP · The Query Is Not the Last Message

Do not confuse with: Vector Search