How I Built Three Apps Without Writing Code: My Journey with Cursor
From a 4-year coding hiatus to launching multiple projects in weeks: How AI-powered development is transforming the way we build software
I finished three projects in 2 months after not coding for 4 years.
This was made possible by Cursor, an AI-based code editor that transformed how I write code - by letting me barely write any code at all. While new AI models get ever more capable, the breakthrough is not just in the models themselves. It's in tools that integrate AI in new, intuitive ways. AI isn't just another feature - it's becoming the core of a whole new generation of software, and we're just seeing the first glimpses. Cursor is exactly that. While AI-assisted coding has been around since GitHub Copilot launched in 2021, it never gained real traction. What's different now is a new wave of development tools that have AI capabilities built into their very core.
What makes Cursor special?
The biggest paradigm shift when working with Cursor is simple - you practically don't write code anymore. Instead, you have conversations about what you want to build. Most of my interactions happen in the Composer window, which acts like a super-powered chat interface. I might type 'Add a function that parses the HTML from a URL and summarizes it using GPT’ or 'Fix this bug where the form submission fails on mobile,' and Cursor will generate, modify, or debug the necessary code across multiple files.
Let me give you a real example. Last week, I needed to add user authentication to a Nuxt.js app. Instead of spending hours setting up routes, forms, and API calls, I simply told Cursor: 'Add Supabase authentication with email and Google sign-in options.' Within minutes, it had created the necessary components, set up the authentication flow, and even added error handling. The implementation worked on the first try and only needed minor enhancements to the user flow.
Of course, it's not a replacement for a human developer. You still need to know what you want to build and handle complex business logic yourself. But that's not the goal - Cursor isn't trying to replace developers, it's augmenting them. Think of it as pair programming with an AI that handles the tedious parts while you focus on the bigger picture.
My personal Cursor experience
Like many technical managers, I started as a data scientist who could build useful things, even if they weren't pretty. This ability to quickly prototype ideas was invaluable early in my career. But as I moved into management, coding fell by the wayside. Every time I tried to build something in my spare time, the overhead of setup and boilerplate code killed my momentum. For four years, I sat on ideas because I couldn't find the energy to bring them to life.
Enter Cursor.
Over Christmas break, I installed Cursor and everything changed. Within two weeks, I built three projects: an AI agent helping me with researching newsletter content, a Google Books/Notion integration that helps me manage my reading list and a small personal dashboard. The game-changer wasn't just coding faster - it was eliminating the mental overhead. Instead of juggling files and remembering where I left off, I could just open my laptop and start describing what I wanted to build.
This momentum carried forward. I've since launched btw-vergleich.de, a website that compares election programs for the upcoming German federal election, and made significant progress on an app idea I've been sitting on for 1.5 years. The barrier between idea and implementation has never been lower.
Cursor at work
Developing personal projects is one thing - working with large corporate codebases is another beast entirely. Multiple developers, strict performance requirements, and complex dependencies all raise the stakes. My personal success with Cursor made me optimistic, but would it hold up in a professional environment?
The answer is a resounding yes. After sharing my experience, a data engineer on my team decided to try Cursor on his backlog. Within a week, he had closed over 20 tickets - everything from adding new features, enhancing existing ones and fixing bugs. He actually complained about 'ticket fatigue' because he was closing tickets faster than he could remember what he worked on.
This wasn't just about speed - the quality held up under review. Cursor-generated code followed our style guides, included proper error handling, and came with comprehensive test coverage. Other teams in the company discovered Cursor around the same time and word spread quickly. Within two months, my entire team had company licenses, and today over 100 people across our company use Cursor daily.
While it's too early to quantify the full impact, the early indicators are clear. Tasks that used to eat up hours - writing test cases, adding documentation, creating boilerplate code - now take minutes. This frees up our developers to focus on what really matters: solving complex business problems and building new features that drive value.
Tips for getting the most out of Cursor
While Cursor is incredibly powerful, knowing a few key strategies can help you avoid common pitfalls and maximize its potential. Here are five tips I've learned from months of daily use:
Focus on the Composer Window Cursor offers both Chat and Composer views, but Composer is where the real magic happens. Unlike Chat, which is for general questions, Composer understands your codebase and can directly modify your files. Skip the Chat and go straight to Composer - simply tell it what feature you want to build or what bug needs fixing.
Enable Agent Mode There's a small but crucial toggle in Composer between normal and agent mode - always choose agent. While it takes longer to generate code, the quality improvement is worth the wait. Agent mode can self-correct and verify its work. For the adventurous, try YOLO mode in settings, which lets Cursor delete files and run terminal commands autonomously.
Commit Early and Often When Cursor goes off track, it rarely happens obviously. Instead, you might find yourself in a cycle of trying to fix increasingly broken code. Regular Git commits are your safety net - when things go sideways, you can easily revert to your last working state and try a different approach.
Start With a Plan While Cursor can build projects from scratch, it works best with direction. Before starting, write a quick project plan: your goals, preferred tools, and key requirements. Run this plan through ChatGPT or Claude first to get a basic prototype, then let Cursor expand it into a full project. I recently used this approach to build a Nuxt.js app with Supabase authentication in record time.
Augment with Documentation Like all AI tools, Cursor sometimes needs help with newer libraries or specific issues. Share relevant documentation or GitHub issues in your prompts - this dramatically improves its ability to generate correct, up-to-date code.
These practices help maintain momentum while avoiding the occasional AI rabbit holes that can eat up your time.
Outlook & Conclusion
While I've focused on Cursor, it's just one player in a rapidly evolving landscape. GitHub Copilot has launched a powerful new version, Cline offers a compelling alternative to Cursor, and specialized tools are emerging for specific needs:
Lovable for browser-based app development
v0 for frontend-focused work
Replit Agent for backend development
The field is moving so quickly that specific tools and features I've described might be outdated in a few months. But that's not what matters. What's happening here is a fundamental shift in how we build software. AI isn't just helping us code faster - it's changing our relationship with development itself. The tedious parts - boilerplate, testing, documentation - are increasingly handled by AI, letting us focus on the creative and strategic aspects of building software.
This is a unique moment. In a year, AI-assisted development will be the norm. But right now, it feels like having superpowers. Whether you're a seasoned developer or, like me, someone who just wants to bring ideas to life, these tools supercharge your productivity.
Don't wait for this to become standard practice. Install Cursor today, try building something small, and experience this shift firsthand. The gap between idea and implementation has never been smaller - take advantage of it.
What has been a gamechanger compared to Github Copilot (which I used the previous year) is that you can
* actively manage the context if needed (add / remove files)
* overwrite existing lines of code instead of just doing smart "autocomplete"
* multi-file edits
As you said: Github is now under pressure to deliver and announced an "agent-mode", but I'm already out of my license. Things are turning so fast
What's your stance on the 70% problem?
https://addyo.substack.com/p/the-70-problem-hard-truths-about