I've put together this guide on 'Agentic Coding' based on my experiences trying to use AI coding agents (like Copilot, ChatGPT, Cursor) effectively for more than just simple code generation.
While AI can generate code quickly, I found it challenging to consistently get reliable, production-quality results in a responsible way. It felt like there was a need for clearer principles and practices beyond just basic prompting techniques. Inspired partly by discussions like Ethan Mollick's 'Speaking things into existence', I tried to structure an approach.
This guide outlines 6 core principles (covering Developer Accountability, Verification, Security, Quality, Human-led Design, and Adaptation) and 28 specific, actionable practices across the development workflow – from preparation and strategic usage to interaction, review, quality assurance, and mindset.
The goal is to provide a framework for developers to leverage the power of AI collaborators effectively while maintaining control, ensuring code quality, and managing risks such as technical debt and security vulnerabilities.
It's hosted on GitHub Pages, the source is written in Markdown, and it's shared under a CC BY 4.0 license, available for anyone to use, adapt, or contribute to.
I'm sharing this here primarily to get feedback from the thoughtful HN community.
* Do these principles and practices resonate with your own experiences using AI for coding?
* Are there crucial aspects or best practices I might have missed?
* Any suggestions for improvement or alternative perspectives?
All feedback is welcome! Thanks for checking it out.
greenbhj•9mo ago
I've put together this guide on 'Agentic Coding' based on my experiences trying to use AI coding agents (like Copilot, ChatGPT, Cursor) effectively for more than just simple code generation.
While AI can generate code quickly, I found it challenging to consistently get reliable, production-quality results in a responsible way. It felt like there was a need for clearer principles and practices beyond just basic prompting techniques. Inspired partly by discussions like Ethan Mollick's 'Speaking things into existence', I tried to structure an approach.
This guide outlines 6 core principles (covering Developer Accountability, Verification, Security, Quality, Human-led Design, and Adaptation) and 28 specific, actionable practices across the development workflow – from preparation and strategic usage to interaction, review, quality assurance, and mindset.
The goal is to provide a framework for developers to leverage the power of AI collaborators effectively while maintaining control, ensuring code quality, and managing risks such as technical debt and security vulnerabilities.
It's hosted on GitHub Pages, the source is written in Markdown, and it's shared under a CC BY 4.0 license, available for anyone to use, adapt, or contribute to.
I'm sharing this here primarily to get feedback from the thoughtful HN community. * Do these principles and practices resonate with your own experiences using AI for coding? * Are there crucial aspects or best practices I might have missed? * Any suggestions for improvement or alternative perspectives?
All feedback is welcome! Thanks for checking it out.