AI First Software Development

Kyle Redelinghuys -

I've been watching all the debates around AI in software development with fascination. While everyone's talking about what might happen, I've been quietly using AI to completely change how I build software. The improvement in development time is dramatic - somewhere between 5-15x faster than before.

The Genesis of AI-Assisted Development

While building GitPilotAI, I noticed my development approach had shifted entirely. These days, I start by collaborating with AI through Cursor, and lately Windsurf, (AI-powered IDEs) to plan architecture, think through edge cases, and design solutions. It's like having someone to bounce ideas off at any time of day, which has been incredibly valuable.

Here's a practical example: I needed to monitor storage on my home media server. Through AI collaboration, I built and deployed a complete solution in about 15 minutes. Nothing fancy - just a service that monitors multiple storage devices, runs when I want it to, and sends email alerts based on thresholds I set. What made this interesting wasn't the complexity, but how quickly I went from idea to working solution.

The New Development Workflow

My process looks quite different now:

  1. Chat through the requirements and edge cases with AI
  2. Use AI-assisted coding (mainly Cursor Composer or Windsurf Cascade) for the bulk of the work
  3. Review and adjust the code where needed
  4. Deploy and set up monitoring

Combined with some solid deployment practices, this gets things into production really quickly.

The Era of Personal Software

This is where things get interesting. We're moving into a time where people can build exactly what they need. The combination of AI assistance and modern development tools means we can create sophisticated, personalised solutions for specific problems.

I've spent a lot of time in fintech, and I'm seeing the same patterns we saw there - moving from generic solutions to highly customised ones. The barrier to entry is still there - you need to know how to code for now. But that's changing quickly as new tools emerge.

Current State and What's Next

The tools are still maturing, and you definitely need to understand software development to use them well. But the impact on both developers and users is going to be significant. Developers can focus on the interesting parts of building software, and users get exactly what they need rather than settling for what's available.

I'm less interested in the ongoing AI debates and more focused on using it to build better solutions right now. It's working well, and it's only getting better.