Skip to content

spec-driven-ai-coding-with-openspec #203

@pushkar-gulkari

Description

@pushkar-gulkari

Title of the talk

Spec-Driven AI Coding with OpenSpec

Description

"Vibe Coding" being a largely known term not only among the dev community but in the overall AI space. People are trying it out, loving it, using it for simple task which is great. When it comes to production grade complex applications, vibe-coding kind of make things complicated as they hallucinate requirements, lose context across sessions, and produce code that looks right but isn't.

We need structure or a framework with which we can have the planning, transparency and clear picture of the path of the development before even AI start making it happen for us. In this session, we'll contrast the vibe coding approach against Spec-Driven Development (SDD) using OpenSpec, a lightweight open-source framework that brings structure to AI-assisted development.

We will explore how OpenSpec bridges human intent and machine execution by turning informal feature ideas into structured proposals, living specs, and traceable tasks — all inside your repo, all readable by any AI tool. If you're already experimenting with AI coding assistants and hitting the ceiling of what a single prompt can reliably deliver, this one's for you.

Table of contents

1. Introduction (4 mins)

  • Quick personal introduction and experience.
  • Briefly introduce the concept of vibe coding and its limitations.
  • Introduce Spec-Driven Development (SDD) and OpenSpec as a solution.

2. Demo 1: Greenfield (11 mins)

  • Start with simple FastAPI app with a single endpoint.
  • Show how to use OpenSpec to create a structured proposal for the feature.
  • Generate the code using AI and show how it aligns with the spec.
  • Highlight the benefits of having a clear spec in guiding AI code generation.

3. Demo 2: Brownfield (11 mins)

  • Take an existing codebase and identify a new feature to add (e.g., user authentication (JWT)).
  • Use OpenSpec to create a proposal for the new feature, considering the existing code structure.
  • Generate the code and demonstrate how it integrates with the existing codebase.
  • Emphasize how SDD helps in maintaining code quality and consistency in a complex codebase.

4. Where This Fits in Your Code Stack (7 mins)

  • Discuss trade-offs between vibe coding and SDD.
  • Discuss why SDD is different than other methods.

5. Conclusion and Q&A (7 mins)

  • Summarize the key takeaways from the session.
  • Open the floor for questions and discussions.

Duration (including Q&A)

40 minutes

Prerequisites

  • Basic understanding of AI-assisted coding and development.
  • Excitement to explore structured approaches to AI coding.

Speaker bio

I am a senior software engineer with over 8+ years of experience in building scalable backend systems. Being someone who got early Python adaption have made me a big fan of the language and its ecosystem. I had worked around Python and related technologies and frameworks like FastAPI, Django, etc. I have been working with AI-assisted coding tools for a while and have seen the evolution of these tools from simple code generation to more complex, context-aware assistants.

I am passionate about sharing knowledge, mentoring engineers, and simplifying complex technical concepts into practical, real-world insights. I try to incorporate and encourage best practices in software development, and I am always looking for ways to improve the development process. I like to connect with people and help them in their journey of learning and growth in the tech industry.

LinkedIn: Pushkar Gulkari

The talk/workshop speaker agrees to

Metadata

Metadata

Assignees

No one assigned

    Labels

    scheduledThe scheduled talkstalk-proposalNew talk of Python Pune meetup

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions