Point Enrichr at any site and it tells Claude exactly how the magic works — GSAP timelines, Framer Motion, live WebGL shaders and textures, pulled straight off the canvas. ~90% accurate on Awwwards-tier builds. enrichr.dev — beta spots open, come break it.
Pulling shader and texture data straight off the canvas is the clever part; that's exactly what screenshots and DOM dumps miss. It pairs well with agents that can already build this tier of site: one Muse Spark build is a single-file marketing site with GSAP transitions, Lenis scroll and Three.js, made from 5-6 prompts with no manual code edits. Feeding Enrichr's breakdown into prompts like those seems like a strong loop. That build and others: https://shipwithmuse.live/categories/apps-and-websites (I help curate it)
This is brilliant - turning animation source-code into readable documentation. The real insight is that most design tools (Figma, Adobe) capture the visual design but not the motion logic. Sites like Awwwards look polished but the motion patterns are a black box unless you inspect the code. A 90% accuracy reverse-engineering tool means you can measure interaction patterns across the best-in-class sites and use that as a design research input. Have you found common patterns emerging - like specific animation libraries or timing models that separate top-tier interactions from mid-tier ones?
Really good writeup, thanks for sharing it. What's the next thing you're planning to try here?
Great question, we will have to wait and see!