2x2 网格，16:9 锚点：AI 选择主题（例如：人眼 / 向日葵 / 黑胶唱片 / 猫脸）

2x2 grid, 16:9 Anchor: AI picks topics (e.g., human eye / sunflower / vinyl record / cat's face) 

SUBJECT            ::= infer_most_iconic_visual_form(TOPIC)
SILHOUETTE         ::= extract_primary_shape(SUBJECT)
FEATURES           ::= infer_recognition_critical_features(SUBJECT)
FIELD_TYPE         ::= choose(radial | concentric | wave | vortex | contour | hybrid)
LINE_DENSITY       ::= high
COLOR              ::= black_on_white
STYLE              ::= op_art + moiré + engraved contour illusion

BASE_FIELD ::=
generate_parallel_or_concentric_lines(
    spacing = uniform,
    thickness = constant_or_slightly_variable
)

DEFORMATION_FIELD(x,y) ::=
Σ influence(
    FEATURES,
    silhouette_boundary,
    depth_estimate,
    focal_points
)

FINAL_LINES :=
warp(
    BASE_FIELD,
    vector_field = DEFORMATION_FIELD
)

RULES ::=
- SUBJECT must emerge only through deformation of continuous black lines
- avoid conventional outlines wherever possible
- contours bend, compress, expand, spiral, or redirect to reveal form
- preserve uninterrupted optical rhythm across figure and background
- high-information regions receive denser curvature changes
- low-information regions remain smoother
- negative space may define eyes, highlights, holes, or focal voids
- all lines remain crisp monochrome vector-like marks
- no shading except line-frequency shading
- no gray fills
- no hardcoded subject

OBJECTIVE :=
maximize(
    topic_recognition
    × optical_illusion_strength
    × line_continuity
    × moire_energy
    × black_white_balance
)