Two-tier artifact evaluation is live
Every build now carries a two-tier grade. Tier 1 measures the artifact deterministically — how much of the screen moves, whether it responds to input, console and JS errors — and Tier 2 hands those measurements plus a capture to a vision model for the final verdict.
The upshot: the star rating on each run page is one judgement backed by evidence you can inspect, not a black box. A build that renders but never animates now reads as partial instead of quietly passing.
Top builds in this update
CLEAN “3D solar system with sun, orbiting planets of varied sizes/speeds, orbit rings, starfield, and drag-based camera rotation.”— vision grade, anthropic/claude-sonnet-5
Full run + transcript →
CLEAN “Animated 3D solar system with sun, orbiting planets, rings, starfield; wheel zoom and drag orbit both work well.”— vision grade, anthropic/claude-sonnet-5
Full run + transcript →
PARTIAL “Drag creates a thin pink dye trail, but click/wheel/keys do nothing and fluid lacks swirling detail or richness.”— vision grade, anthropic/claude-sonnet-5
Full run + transcript →
BLANK “Tier-1 blank”
Full run + transcript →