Blog

13 October 2026 · research · development

A Letter on Scoring Changes — 13 October 2026

Classification is the same. Kinship tags now share a room, so Sad can meet Melancholic. Meaning returns to 1× so we can isolate the effect.

As of 13 October 2026 we have changed how TEV scores classified tags. Classification itself has not changed. A query or a source still receives the same five-axis labels, from the same dictionaries, with the same limits. What changed is how two of those label lists are compared.

Until today, overlap was binary and equal: a tag either matched or it did not. Sad against Melancholic scored as a miss. That was too harsh on the axes where several tags share a conceptual space.

Two kinds of axis

You already know the five categories. They also split another way.

Exact axes — Genre and Scale

These name a world and a size. Different readers tend to land on the same label. A product launch is Sci-Fi/Futuristic and Medium, not Abstract and Intimate, even if the theme mentions a wrist. Overlap here stays binary.

Kinship axes — Emotion, Energy, and Meaning

These are vaguer and more interpretive, so a near miss should still contribute to the score. Axes remain independently scored. We do not want to punish a Sad query because the card was labelled Melancholic.

The new approach

Inside each kinship axis, tags live in disjoint rooms — concept sets. One tag, one room. No tag sits in two rooms. Genre and Scale have no rooms.

When we score a query tag against a card:

  • same tag → 1.0 (exact)
  • different tag, same room → 1 / size of that room (kin)
  • different room, or a Genre / Scale miss → 0

The query is not rewritten. Sad stays Sad. A card tagged Heartbroken can still rank, at 0.25, because both sit in the same room.

Sad against Sad is 1.0. Sad against Heartbroken is 0.25. Sad against Happy is 0.
Sad against Sad is 1.0. Sad against Heartbroken is 0.25. Sad against Happy is 0.

We are also returning Meaning to a 1× weight — as it was before the Cohort 02 developments. I hope this does not disturb anyone’s plans. I want to isolate the effect of rooms, so every axis needs a clean, normalised state. Divya, with Axis Tilt, will likely bring more dynamic category weighting when the time is right.

The new concept sets

These are the rooms now in force. The names on the diagrams are only for reading; the engine stores the tags, not the room titles.

Emotion

Emotion: grief, lift, poise, alarm, jolt.
Emotion: grief, lift, poise, alarm, jolt.
  • Sad, Melancholic, Heartbroken, Depressed
  • Happy, Ecstatic, Excited, Hopeful
  • Calm, Peaceful, Nostalgic
  • Anxious, Fearful, Angry
  • Surprised, Confused

Energy

Energy: stillness, menace, surge, disorder.
Energy: stillness, menace, surge, disorder.
  • Static, Serene, Contemplative
  • Tense, Ominous, Violent
  • Energetic, Dramatic, Lyrical
  • Random/Chaotic, Ludicrous/Absurd, Playful, Euphoric

Meaning

Meaning: hold, passage, care, friction, bond, double. Weight is 1× again.
Meaning: hold, passage, care, friction, bond, double. Weight is 1× again.
  • Retreat, Confinement, Suspension
  • Pursuit, Emergence, Ascent, Threshold
  • Gatekeeping, Stewardship, Witness
  • Contention, Intrusion, Erosion, Fall
  • Return, Recognition, Communion, Agency
  • Paradox, Juxtaposition

How this changes search

Phrase search and card-similar (reverse search) both rank by this score. Results with score 0 drop out.

Each result still returns:

  • score — the number we rank on
  • categories — every tag on the card
  • matched_categories — only the tags that contributed to the score
  • query_categories — the original classified or posted tags, not expanded to the whole room

Classification, ingest, and the five vocabularies are the same as yesterday. We only stopped treating neighbours as strangers.

—

Asycd — Cohort 02. 13 October 2026.