Another small case where an AI coding agent helped me clean up a Framer interaction.
I had a navigation component that needed to stay transparent on image-heavy hero sections, become solid on light sections, and still behave correctly when the mobile menu opens.
With native Framer variants, this quickly turned into too many states, scroll triggers, and page-specific fixes.
So we moved the logic into a small override:
— toggle the navigation background
— switch menu item colors
— handle mobile open/closed states
— keep the behavior reusable across pages
This wasn’t a “click one button and AI makes it beautiful” story.
Honestly, it was almost the opposite. I had to break down the logic myself: when the background should appear, when it should disappear, which elements should use the new color token, and how everything should behave across different states.
After a few iterations, the agent helped turn that logic into a working override.
To me, this is where AI feels useful in design and no-code right now. Not as magic, but as implementation leverage when you already understand what you’re trying to build.
There is still friction between idea and result. But it feels like that gap is getting smaller every month.
Another small case where an AI coding agent helped me clean up a Framer interaction.
I had a navigation component that needed to stay transparent on image-heavy hero sections, become solid on light sections, and still behave correctly when the mobile menu opens.
With native Framer variants, this quickly turned into too many states, scroll triggers, and page-specific fixes.
So we moved the logic into a small override:
— toggle the navigation background
— switch menu item colors
— handle mobile open/closed states
— keep the behavior reusable across pages
This wasn’t a “click one button and AI makes it beautiful” story.
Honestly, it was almost the opposite. I had to break down the logic myself: when the background should appear, when it should disappear, which elements should use the new color token, and how everything should behave across different states.
After a few iterations, the agent helped turn that logic into a working override.
To me, this is where AI feels useful in design and no-code right now. Not as magic, but as implementation leverage when you already understand what you’re trying to build.
There is still friction between idea and result. But it feels like that gap is getting smaller every month.