Redesigning Websites with GPT-5.5 & Images 2.0

By Lukas Margerie

Categories: Design, Product

Summary

GPT-5.5 combined with Images 2.0 eliminates weeks of design work by generating production-ready landing pages in minutes. Creators can now redesign websites through simple prompts and image regeneration, reducing multi-tool workflows to a single AI-powered environment.

Key Takeaways

  1. Use the 'taste skill' in Codex to automatically redesign generated websites, adding interactive elements like sticky scroll, hover effects, and Bento grids without manual coding.
  2. Feed GPT's Images 2.0 output back into Codex for iterative refinement—regenerate specific sections based on landing page context, then replace components to improve overall design cohesion.
  3. Use annotation commands in Codex to delete or modify specific UI elements in real-time (under 1 minute), eliminating manual editing and reducing design iteration cycles.
  4. Compare model outputs (Claude Opus vs GPT-5.5) at identical intelligence settings to benchmark design quality—GPT-5.5 produced layouts that felt less 'AI-generated' with better visual hierarchy.
  5. Complex configurators and interactive tools that previously required weeks of stitching multiple tools together can now be generated entirely within Codex using single prompts.

Topics

Transcript Excerpt

Hey guys, now GPT 5.5 is finally here. And with this great leap forward from OpenAI, this insane next step towards the inception of AGI, I really don't feel limited by what this model can do. I actually feel limited by what I can imagine. And just to show you quick little example of what you can build with this. This is an example built by Pietro Schirano, who's the founder of Magic Path. He built this time machine where you can prompt out the moment in time where you want to go, and Codex using...