A Simple Request Sparks a Technical Challenge
While helping my son prepare for school, we encountered a surprisingly frustrating limitation on Amazon. He wanted to share the contents of his shopping cart with a friend to ensure they were buying identical supplies. Surprisingly, Amazon offers no native way for personal accounts to share a cart—a feature restricted solely to their business platform. The existing workaround, using 'Lists,' is far from intuitive for a child who has already finished building a cart.
Observing this inefficiency, I wondered: could I use the power of modern AI—specifically "vibe coding"—to build this missing link myself?
The AI-Driven Development Process
Without deep expertise in software development, I turned to ChatGPT to bridge the gap. By describing the problem in natural language, I sought a tool that could extract product data from an Amazon cart and encode it into a shareable link. The goal was to create a lightweight Chrome extension that would allow a recipient to see exactly what was in the original cart without compromising privacy or security.
The resulting prototype, which I named "Cart," was impressive. ChatGPT successfully handled the interface design, site publishing, and the construction of the Chrome extension. Within hours, I had a functional-looking product that promised to streamline the sharing process.
The Reality Check: Where AI Meets Corporate Infrastructure
The project served as a humbling reminder that a polished interface is not the same as a functional product. While the link-sharing aspect worked perfectly, the "one-click" transfer to a recipient's Amazon cart failed. When I attempted to use encoded URLs to populate a cart, Amazon’s system rejected the request with a 500 error, as it does not support such direct session manipulation for consumers.
«AI can produce a polished interface before the underlying behavior has been properly tested. The buttons, colors and copy make the project feel finished, but a convincing website is not proof that the feature behind it works.»
Lessons in Prototyping
I had to pivot the project to reflect these technical boundaries. I replaced the broken transfer button with a status notice that clarifies the tool's current limitations. Users can now use it to compare items, but full-scale automation remains blocked by Amazon's technical architecture.
Looking ahead, there are two paths for improvement:
- Browser Automation: The extension could simulate user clicks to add items to the recipient's cart, though this approach is fragile and highly dependent on Amazon's UI layout.
- Simplified Comparison: The current model, which focuses on displaying a curated list for easy comparison, is already a significant improvement over taking screenshots or managing complex registries.
While my prototype isn't a replacement for a native Amazon feature, the process proved that modern AI allows us to turn a simple, everyday annoyance into a tangible technical experiment. The journey from an idea to a working prototype highlights that while the "vibe" of building is easier than ever, the real engineering challenge lies in navigating the complex constraints of large-scale platforms.