GPT Direction — Registries-Pivot Auto-Label Threshold Clarification

Date: 2026-05-30 Reviewer: GPT Council

Clarification from user

The previous wording “if child_count > 50” can be misread as “wait until exactly/over 50 before classification matters.” That is wrong.

Correct principle:

• 50 is a maximum ungrouped display threshold, not a target.
• If a list is not already classified/grouped and it is becoming too long to inspect safely, the classification/grouping process must start immediately.
• Some lists may already be classified; if so, reuse the existing classification/labels rather than adding another grouping layer.
• The purpose is to prevent long flat lists and keep every layer short, inspectable, and scalable.

Revised rule

For each node/list:

1. If existing classification/label/group dimension exists, use it first.
2. If no classification exists and child_count is above the configured safe threshold, trigger CLASSIFICATION_REQUIRED.
3. The maximum default threshold is 50 ungrouped rows, but the backend may choose a smaller threshold by species/list type.
4. Never wait for the list to reach exactly 50 as a goal.
5. Never solve long lists by frontend pagination only; pagination may help UI but does not replace semantic grouping.
6. Labels/grouping dimensions must be PG-backed, governed, pivot-countable, and themselves registered.

Design consequence

The Registries-Pivot design must include:

• classification_status;
• classification_dimension;
• label_ref;
• grouping_required;
• grouping_reason;
• max_ungrouped_threshold;
• suggested_next_grouping;
• classification_workflow_trigger.

This clarification must be incorporated into the next Registries-Pivot macro before acceptance.