10-second screen recording in, working website out: a hands-on test of Ling-3.0-flash-VL's video-to-code
nikola_mr64990 · x · 2026-09-21
@AIjacksaku recorded 10 seconds of scrolling a designer's personal site and fed the raw video — no source code, no long prompt — to Ling-3.0-flash-VL (Ant's Ling-3.0-flash with vision understanding and a visual agent added) to see if it could rebuild the page from video alone.
Verdict: roughly 70% similar by eye — layout, color palette, and overall feel captured; buttons clickable, interactions functional, a decent first draft. The remaining 30% shows: fonts off, interactions stiff, visible gaps side-by-side. Not shippable as-is, but a solid starting point. The author's takeaway: if you have a page in your head but can't describe it, record a similar site and let the model build v1. Officially claimed capabilities (STEM, docs, medical reports) were not tested.
More from coding & agent
- "Don't vibe code if you can't code without AI" sparks skill-dependency debate — prasenx · 2026-09-21
- Model grades itself full marks for a chart that never existed — memokris · 2026-09-21
- AI finds key-independent collisions in six major hash functions — thomasahle · 2026-09-21
- 7 document parsing APIs tested on 11 PDFs: no single winner — latentnoise_ · 2026-09-21
- Building a Claude API chatbot with live web search: 5 open engineering questions — angel-queen123 · 2026-09-21
- Hosted MCP for Odoo launches with OAuth 2.0 and 400 free calls — felix010301013 · 2026-09-21