Claude skips the training set: HTML/CSS rendered in a headless browser makes his site images
shashib · x · 2026-09-09
The author argues "AI-generated images" are two different things: diffusion models reconstruct from patterns learned across billions of scraped photos, versus code-written pages rendered by a browser with no photo dataset involved. His site's hero images and LinkedIn cards use the second kind: Claude writes Python and HTML/CSS against a fixed design system, a Playwright headless browser renders it, and PIL downsamples the capture. No training images anywhere in the path; open-ended illustration is a different matter.
More from coding & agent
- Dev builds Capi, a kids' emotion-story app, in a weekend with GPT-6 Astra and fal Agent — noahsolomon · 2026-09-09
- Browser Demo Renders 139k Trees and 240 Buildings in 4k Characters — tomjohndesign · 2026-09-09
- ZeroClick launches Agent Storefronts, letting businesses sell to AI agents via x402 and USDC — kleffew94 · 2026-09-09
- How to build a self-healing CI pipeline with agents — AI triage, human gate — Pavan_Belagatti · 2026-09-09
- Hand An AI Agent A setup.md On A New Box And It Installs Everything For You — draginol · 2026-09-09
- Netlify demos OpenCode for comparing AI models, costs and task fit — thisiskp_ · 2026-09-09