Clarify the public AI story for Morphēa

[swernerx]

Goal

Turn the AI story into a clearer public-facing narrative with concrete artifacts: what uses AI, what stays rule/gate based, and why this is safer for editable SVG output than a black-box conversion.

Why this matters

"AI vectorizer" is a crowded and often vague category. Morphēa has a more interesting angle: local AI-assisted shape reconstruction with reports, gates, reviewed pseudo-labels, and own-model training evidence. That should be visible without making inflated claims.

Proposed work

• Expand README or site copy with a short "AI with receipts" section.
• Add a simple diagram of the AI-assisted loop: generate/train -> vectorize -> harvest -> review -> compare/gate -> retrain.
• Add links from the homepage/site to docs/ai-in-morphea.md.
• Add one artifact-driven example once a new curated case is stable: input image, Morphēa SVG, report highlights, and model/review evidence.
• Keep language honest about optional MLX/SAM dependencies and current research-stage limits.

Acceptance criteria

• A newcomer can answer: "What AI does Morphēa use today?" within one minute of reading the README/site.
• The copy distinguishes model-assisted ranking/segmentation/training from black-box generation.
• The docs name the current AI surfaces: MLX primitive classifier, raster-target model, optional MLX/SAM segmentation, reviewed pseudo-label learning, and synthetic training data.
• No claims imply universal image vectorization or guaranteed production quality.

Notes

This is a marketing and documentation issue, not a request to add a bigger model. The main value is clarity and trust.