The future of software development is AI-driven coding loops where humans guide high-level design and AI handles the rest. Within the next two years, AI will likely write code better than any human, and we’ll rarely need to look at the code it produces.
In this talk, I’ll share my experience building huddlz, a full-featured web app built mostly by AI with Elixir, where I personally write less than 1% of the code. I’ll also talk about my open-source Elixir implementation of cucumber, written 100% by AI, which serves as an interface to communicate behaviors with the AI in plain English.
To manage challenges and complexity, I developed a structured multistep process that creates self-improving loops, iterating on AI prompts and code while continuously learning and documenting. The goal is for AI to progressively take over implementation with minimal intervention, enabling rapid and creative development.
This talk is geared toward developers excited about AI-driven software development and eager to get the most out of AI today.
Key Takeaways:
- Discover how to establish self-improving AI loops that minimize manual coding and keep AI on task.
Target Audience:
- This talk is geared toward developers excited about AI-driven software development and eager to get the most out of AI today.