[raelovejoy]

What this adds

This captures a distinction that emerged from the existing serialization / AI-assisted-expression work:

• a structure can be internally coherent without being externally legible;
• translating a dense model into a linear or simplified representation can lose important relationships;
• clarity and fidelity are related but not identical.

Canonical change

concepts/legible-systems.md now distinguishes:

• structural coherence
• representational legibility
• translational fidelity

It adds a layered explanation pattern: doorway → working model → inspectable depth, and extends the design test to ask whether simplification preserves relationships that materially affect meaning.

Local application

concepts/legible-ai-assisted-expression.md now connects its existing “writing as a serialization task” section to the broader Legible Systems principle. It explicitly notes that an expression bottleneck can be a representational-legibility problem rather than an understanding problem.

Scope

This PR intentionally does not copy the same material into UCF, the coherent-systems living paper, or project repositories. Those can reference or transform the canonical principle later if a concrete need appears.

No merge requested; this remains available for review.