Welcome back to our series on how Propel Ventures is pioneering the use of Generative AI into software development to boost productivity and enhance quality. In our previous blogs, we discussed the foundational steps and early successes in adopting Generative AI for requirements development and generating code for UI components. Today, we delve deeper into our journey, sharing insights from our recent experiments with GitHub CoPilot and the transformative impact they've had on our development process.
The Evolution of AI in Development at Propel
At Propel, our commitment to innovation has led us to explore the vast potential of off-the-shelf AI products in software development. Our experiences have been a mix of challenges and triumphs. Initially, while these AI tools showed promise in improving outcomes, their rigidity and the significant effort required for customization presented hurdles. Despite these obstacles, our pursuit of efficiency and quality has borne fruit, particularly with the use of GitHub CoPilot.
GitHub CoPilot: Living up to the hype
A natural choice to begin our code generation journey was GitHub CoPilot, and we were delighted to discover two of its extensions to be incredibly valuable.
1. The IntelliSense-like Extension: This early tool offered a glimpse into the future, providing code suggestions based on comments or the project's context. It proved invaluable for autocomplete functionalities, streamlining the coding process.
2. The Copilot Chatting Extension: The true breakthrough came with this second extension. It transformed our interaction with AI, allowing us to communicate more effectively with the copilot. This extension has been instrumental in generating high-quality unit tests for API controllers, business logic, and other complex components of our software, enhancing both efficiency and quality.
Test Driven Development with AI
One of the most exciting aspects of our AI journey has been challenging traditional software development methodologies. One area where CoPilot proves its value is by transforming the conventional approach of writing code before tests and transitioning to Test Driven Development. Although Test Driven Development is not a new concept, it is rarely implemented in practice. However, with the aid of Generative AI, Test Driven Development becomes more natural and intuitive. By establishing unit tests or contracts as a priority and then utilizing AI to generate code that meets these specifications, we envision a software development process that is both more efficient and higher quality.
Propel's AI-Powered Code Generation Extension
Inspired by the success of these GitHub extensions, we embarked on creating our own VS Code extension for code generation. By focusing on usability features like context menu integration and AI-generated file outputs, we've lowered the barrier to adoption for our team, facilitating a smoother transition from prototypes to production.
Looking Forward
As we continue to explore the capabilities of Generative AI, our goal is to redefine the norms of software development. By flipping traditional methodologies on their head and embracing AI's potential, we believe we can unlock increasing levels of productivity and quality in our projects.
Stay tuned for more updates as we further our journey into the future of software development with Generative AI. Propel Ventures remains at the forefront, pushing the boundaries of what's possible and sharing our learnings with the wider community.
Thank you for following our series. Your engagement and feedback inspire us to keep innovating and sharing our experiences. Together, let's shape the future of software development.
---
*This blog is part of a series exploring Propel Ventures' use of Generative AI in software development. Be sure to read the first two entries (Our Experiments with User Story Generation and Our Experiments with UI Code Generation) for more background and insights into our journey.*
