I, a non-coder, built a LinkedIn AI copywriting app in a week

Hey HN, I'm a HR professional who's always dreamed of building software but lacked coding skills. My last attempt at creating a website ended with a $10k bill and a clunky Wordpress mess. But last month, after a year of tinkering with GenAI, I decided to tackle a problem I personally faced: writing engaging LinkedIn posts.

My challenge was to build a web app that writes in the style of successful LinkedIn creators. All without knowing how to code.

Here's the journey:

- Starting point: A basic React template from Vercel and a Figma mockup. - AI assistance: I turned to Claude by Anthropic. I'd upload screenshots of my designs and ask, "Act as a UI/UX expert. Rate this design X/10. How can I make it 10/10?" This iterative feedback loop refined my designs quickly. - From mockup to code: I created a Claude Project, uploaded my template files, and asked for working code based on my mockups. This gave me a foundation, but the real work was just beginning. - Debugging: Every error became a puzzle. I'd copy the error to Claude, get a fix, implement it, and repeat. This cycle taught me more about coding than any tutorial could. - A pleasant discovery:Cursor IDE, a VSCode-like editor with integrated AI. Suddenly, I could ask for code changes, explanations, and bug fixes without context switching. The speed of iteration went up like 50%.

This is the final result: www.postin.ai - a functional web app with user authentication, payments, and AI-powered post generation, built in under two weeks.

But here's what surprised me most: - The power of specific prompts: Asking Claude to "Act as a senior React developer reviewing my code" yielded much better results than generic questions. - The importance of understanding data flow: Even with AI, I needed to grasp how data moved through my app to debug effectively. - The unexpected complexity of state management: What seemed simple in mockups became a tangled web of React states and effects. - The challenge of API integration: Connecting to Anthropic's API and handling rate limits taught me more about async programming than I ever expected to learn.

This experience has me questioning the future of coding. Will AI lower the barrier to entry for software development, or will it simply shift the skills needed from syntax memorization to effective prompt engineering and system design?

I'm still learning and iterating. Currently working on a calendar view for post scheduling, which is proving to be an interesting challenge in date manipulation and UI state management.

I'm curious to hear from the HN community:

1. How do you see AI changing the landscape for non-traditional developers? 2. For the experienced devs: What crucial areas of software development am I likely overlooking in this AI-assisted approach?

Happy to dive deeper into any aspect of this journey. Let's chat!