Eleanor Berger discusses the transformative impact of AI-assisted coding, emphasizing the need for clear, detailed instructions and tailored prompting to effectively integrate AI agents into software development workflows. She highlights her project Ruler for managing AI tool configurations, shares her preferred AI models, and advises developers to adopt new habits like detailed prompting and voice dictation to maximize AI productivity.
In this discussion, Eleanor Berger, an AI engineering consultant and educator, shares insights into AI-assisted software development and her comprehensive course on the subject. She highlights the growing demand from software teams and leaders for guidance on integrating AI into coding workflows, especially as newer AI models have significantly enhanced capabilities. Eleanor emphasizes the transformative potential of delegating tasks to AI agents that can operate autonomously in the background, improving productivity and code quality without constant human intervention.
Eleanor discusses the importance of clear and detailed specifications when working with AI coding agents. Unlike humans, AI models do not ask for clarifications and will attempt to complete tasks even if instructions are vague, which can lead to unintended results. She advises providing comprehensive instructions, including constraints and what not to do, to ensure the AI produces the desired output. She also notes that different AI models may require slightly different prompting styles, with some needing more explicit repetition or emphasis, but the core principle remains clear communication.
The conversation touches on Eleanor’s project, Ruler, which addresses the challenge of managing rules and configurations for AI tools across diverse developer preferences and environments. Ruler centralizes context and configuration for multiple AI coding agents, facilitating consistent behavior across teams. This project exemplifies Eleanor’s approach of leveraging AI to build AI tools, as she developed Ruler entirely through AI-generated code based on her specifications. She also discusses the emerging standard of agents.md files for defining agent rules within repositories, which helps unify and streamline AI agent behavior.
Eleanor shares her preferred AI models for coding tasks, favoring GPT-5 for complex and agentic work, GPT-5 Mini for simpler tasks, and Grock Code Fast for speed despite its less polished behavior. She stresses the value of developers investing time to understand their chosen AI tools, comparing it to learning a programming language or compiler, as this knowledge enhances control and efficiency. She also prefers using command-line tools over MCP (Machine Control Protocol) servers for AI interactions, as they are familiar, transparent, and easier to manage.
Finally, Eleanor offers practical advice for those starting with AI-assisted coding: use more words and provide as much detail as possible in prompts. She finds that dictating instructions aloud helps her articulate fuller, clearer prompts, which AI models handle well even if the input is unstructured. This habit reduces errors and improves outcomes. Eleanor encourages developers to embrace new workflows and habits, such as voice dictation and detailed prompting, to fully leverage AI’s capabilities in software development. She invites viewers to explore her course and newsletter for ongoing tips and guidance in this rapidly evolving field.