I'm not a senior developer. I know my way around Next.js and I've built a few projects, but I wouldn't call myself experienced with Framer Motion, Tailwind v4 or the kind of polished component architecture you see in products like ShipFast or Supastarter.
A few days ago I decided to try something: use AI as a coding partner to build and ship an actual product — not a tutorial project, not a clone, something I could sell.
The idea was simple. AI startup landing pages all look the same, and most developers rebuild the same components from scratch every time. A kit of production-ready components seemed useful and achievable.
How the session went
I started by describing what I wanted — a landing page kit for AI startups, dark mode, animated, Next.js. AI helped me think through the positioning first: what already exists, where the gap is, what price makes sense.
Then AI built the kit component by component. I'd describe what I wanted, AI would generate the code, I'd drop it into my local dev server and tell it what was broken. There were real bugs — a scroll hijacking issue in the Hero section, a toggle cursor that animated in the wrong direction in the Pricing table, grid layout issues in the Bento component. Each one took a few back-and-forth messages to fix.
So I wasn't just passively accepting output — I was testing, observing, and directing.
What got built
Eight components in total:
- Navbar — sticky, blur backdrop, animated mobile menu
- Hero Split — left copy, right live AI chat demo with streaming animation
- Logo Cloud — infinite marquee with hover color reveal
- Bento Grid — asymmetric feature grid, scroll-triggered animations
- Testimonial Marquee — dual row, opposite directions
- Pricing Table — monthly/yearly toggle, 3 plans
- FAQ Accordion — two columns, JSON-LD schema for SEO
- CTA Section — layered glow background, email capture
All TypeScript strict, Tailwind v4, Framer Motion. Mobile responsive.
Conclusion
I didn't write the code and I wouldn't pretend otherwise. What I did was test every component, catch every bug, configure the tooling, write the Gumroad page, and ship it.
Is it my code? No. Is it my product? Partly.
But it works and it solves a real world problem.
Do you think that's a legitimate way to build things in 2026?
Let me know in the comments ;)
Live demo: https://ai-landing-kit-two.vercel.app/
Gumroad: https://valentino760.gumroad.com/l/ai-landing-kit
Top comments (0)