Upgrade to Pro — share decks privately, control downloads, hide ads and more …

The New Developer Workflow: How AI Transforms I...

The New Developer Workflow: How AI Transforms Ideas into Code

Modern application development is evolving rapidly with the help of generative AI. This session explores a practical, end-to-end developer workflow—from idea generation to production-ready code—powered by tools like GitHub Copilot, ChatGPT, and model customization via MCP Servers.

It begins with using conversational AI models to research and refine product ideas, continues with building lightweight MVPs using tools like v0.dev or Lovable.dev, and transitions into guided code generation, documentation, testing, and reviews—all driven by AI models tuned to specific developer needs.

Attendees will learn how to:
• Use AI tools like ChatGPT and Gemini for deep product research and feature planning
• Leverage GitHub Copilot Custom Instructions and MCP Servers to match team coding styles
• Generate reusable prompts for consistent code and test generation
• Integrate Copilot into everyday development workflows using IDEs like VS Code or Cursor
• Automate testing, documentation, and even code reviews
• Understand which AI models (GPT-4o, Claude 3.7, Gemini, etc.) are best suited for specific development tasks

The session highlights how developers can build smarter and faster by treating AI not just as a code assistant, but as a collaborative teammate throughout the entire software development lifecycle.

Daniel Sogl

April 30, 2025
Tweet

More Decks by Daniel Sogl

Other Decks in Programming

Transcript

  1. The New Developer Workflow How AI Transforms Ideas into Code

    Daniel Sogl @sogldaniel Software architect
  2. Daniel Sogl • Software architect @ Thinktecture AG • Focus:

    Angular, Capacitor and AI-Coding • Creator of https://codingrules.ai • Socials: https://linktr.ee/daniel_sogl About me
  3. From suggestions to end-to-end support Idea MVP Code Review •

    Ideas can be explored and refined with AI tools like ChatGPT or Gemini. • Product concepts can be validated using MVP platforms like Lovable.dev or Firebase Studio. • Code can be generated efficiently with AI- powered plugins like GitHub Copilot or AI-first IDEs like Cursor and Windsurf. • Code can be tested, documented, and reviewed using AI agents like GitHub Copilot and Google Gemini Code Assist. The New Developer Workflow
  4. • Start with a rough idea – e.g., a productivity

    tool or developer helper • Use ChatGPT or Gemini Deep Research to refine your concept, identify use cases, and generate user stories • Define clear features, personas, and goals • AI is your thinking partner – not just your coding assistant Tipp: Use GPT Researcher as an Open-Source alternative with your own API keys Finding the Idea: Deep Research with ChatGPT
  5. • Use platforms like v0, Lovable.dev, or Firebase Studio •

    Generate first screens, forms, flows, and ideas • Visualize user interaction and validate core features • Outcome: A lightweight MVP that guides implementation Build a Visual MVP with AI Tools
  6. • Create a GitHub repository with your app structure (Angular

    or another framework) • Define the initial structure: folders, components, services • Add a task list and define your scope (start small!) • Keep it simple – Copilot and AI tools thrive in structured environments Tipp: Use GitHub Copilot MCP-Server to create the tasks based on your previously created idea Set Up Your Project Foundation
  7. • Configure Custom Instructions to match your coding style (with

    the help of codingrules.ai) • Example: How functions, tests, and comments should look or how the architecture is defined • Set up the MCP Server to fetch external knowledge • AI works best when it knows your rules – make them explicit Tipp: use short, self-contained statements Customizing Copilot with Instructions & MCP
  8. • Break down the app into clear tasks: services, components,

    logic units (sequential-thinking MCP server) • Prepare reusable prompts for repeated patterns (e.g. “Generate an Angular Service that...”) • Reuse your instructions to guide Copilot consistently • Save time and ensure quality with predictable outputs Define Smart Tasks with Reusable Prompts
  9. • Use Copilot with VS Code or AI-first IDEs like

    Cursor or Windsurf • From prompts and comments to full services or UI components • See how Copilot auto-completes entire flows – respecting your structure (using Agent mode) • Your job: guide, validate, refine Generate Code with GitHub Copilot
  10. Bonus: Choosing the Right AI Model • Balance between cost

    and performance: Go with GPT-4.1, GPT-4o, or Claude 3.5 Sonnet • Fast, lightweight tasks: o4-mini or Claude 3.5 Sonnet are your buddies • Deep reasoning or complex debugging: Think Claude 3.7 Sonnet, o3, or GPT 4.5 • Multimodal inputs (like images): Check out Gemini 2.5 Pro or GPT-4o
  11. • Let Copilot write unit tests based on the logic

    it just generated (or the other way around using TDD) • Improve readability by generating inline comments and doc blocks • Bonus: Generate Markdown files like README.md or api.md • Result: More time for logic – less for boilerplate Write Tests and Documentation with Copilot
  12. • Use GitHub Copilot for inline code reviews • Let

    it highlight performance issues, refactoring suggestions, or missing checks • See suggestions directly in VS Code • Copilot becomes your pair reviewer – early and fast Review Code in Your IDE with Copilot
  13. • Let Copilot review your Pull Requests on GitHub •

    It generates meaningful comments on diffs and suggests improvements • Auto-generates PR descriptions summarizing your changes • You focus on logic – AI helps with structure and clarity Review PRs with Copilot on GitHub
  14. • We started with an idea... • Turned it into

    a visual MVP • Implemented logic with Copilot • Generated tests, docs, and reviews From Idea to App – The Result Idea MVP Code Review
  15. • Clear prompts = Better suggestions • Customize Copilot for

    better results • Use MCP-Servers for external context • Always validate AI-generated code • Treat AI as a partner, not a replacement Lessons Learned & Best Practices