Visual Studio Code hosts Pamela and Eleanor Berger as they discuss advanced tactics for AI-assisted coding with Copilot, agentic workflows, and open-source tools, providing practical advice for developers.

Taming AI Assisted Coding Models with Eleanor Berger

Pamela and Eleanor Berger dive into the art of AI-assisted coding, focusing on practical strategies and hands-on workflows for developers using Visual Studio Code and GitHub Copilot.

Key Topics Covered

  • Delegating to Agentic Models:
    • The rationale for offloading coding tasks to background AI agents and how this changes daily development.
    • Overview of different agent tooling including the agents.md standard for documenting agent behavior.
  • GitHub Copilot and Actions Integration:
    • Using Copilot for code completion, chat, and automation within VS Code.
    • GitHub Actions for automating workflows and background tasks.
  • Writing Specs Models Actually Follow:
    • Techniques for drafting effective specifications that LLMs can interpret accurately.
    • Strategies for working with multiple LLMs (GPT-5, Mini, Grok) and aligning output with developer intent.
  • Sandboxed Evaluations and Guardrails:
    • Running evaluations in sandboxes to safely test AI-generated code.
    • Implementing guardrails to minimize errors or unintended outcomes.
  • The Ruler Project and Centralized Rules:
    • Introduction to Eleanor’s open-source Ruler project for managing rules across agentic workflows.
    • Benefits of centralizing AI governance for teams and individual developers.
  • Workflow Innovations:
    • Designing CLI-first approaches alongside interactive voice controls in VS Code.
    • Managing multiple agents and workflows for complex coding scenarios.
  • Model Face-Offs and Personality Steering:
    • Comparing the capabilities of different AI models and learning to steer their unique behaviors.

Practical Advice

  • Create richer, more targeted prompts to get better AI code assistance.
  • Use agents.md as a standard for documenting and communicating agent behavior within your codebase.
  • Test AI-generated code in isolated sandboxes before merging into production.
  • Employ centralized governance (like Ruler) to maintain consistency and safety in multi-agent environments.
  • Explore voice and CLI tools for diverse interaction modes within VS Code.

For more episodes, subscribe to the VS Code Insiders Podcast on your favorite platform.

Pamela and Eleanor Berger provide a candid, technical deep dive into AI-assisted coding, highlighting practical tools, standards, and open-source innovations for modern developer workflows.