Episode 18 - Vibe Modeling: The Next Evolution Beyond Coding

Episode Summary

Martin and Adam explore the future of software development where AI handles implementation while humans focus on collaborative modeling. They distinguish “vibe coding” (letting AI write code from vague descriptions) from “vibe modeling” - using event modeling sessions to precisely define system behavior before any code exists. The discussion positions event modeling as the bridge between human intent and AI implementation.

Main Discussion Points

• Vibe Coding vs Vibe Modeling: The critical difference between letting AI improvise implementations versus using AI to implement precise specifications
• Collaborative Definition: How event modeling sessions bring stakeholders together to define exact system behavior
• The Future of Development: Developers shifting from coding to facilitating modeling sessions and validating AI-generated implementations
• AI as Implementation Engine: Using AI to generate code from well-defined event models rather than vague requirements
• Precision in Specifications: Why precise specifications through event modeling eliminate AI hallucinations
• Event Sourcing’s Future: How event sourcing becomes even more valuable when AI handles routine implementation tasks

Key Takeaways

The future of development separates collaborative modeling from implementation mechanics. “Vibe modeling” through event modeling sessions produces precise specifications that AI can reliably implement, while “vibe coding” from vague descriptions produces unreliable results requiring extensive debugging. Event modeling provides the structured format that bridges human business understanding and AI code generation. Developers evolve from code writers to facilitators who guide modeling sessions, validate specifications, and review AI-generated implementations against those specifications.

Memorable Quotes

1. “Vibe modeling. Yes. There it gets really really interesting. So it kind of goes into maybe even uh some of the practices that are out there already like mob programming and pair programming” - Martin Dilger
2. “Coding is nothing that that amazes me. You can just generate code. It’s it’s really generating code is not difficult. It’s no longer the it’s no longer the interesting part. It’s really like you said the understanding of the solution, right?” - Martin Dilger
3. “The life cycle of a slice is really within the context of what the slice is part of. So I rarely see, you know, one single uh slice being approved or disapproved” - Adam Dymitruk
4. “Slices are very copy paste. So even the UI ends up getting copy pasted” - Adam Dymitruk
5. “The work is done. My subjectivity has no place” - Adam Dymitruk

Key Learnings

1. Vibe coding without understanding is dangerous, but vibe modeling with event modeling provides valuable guard rails
2. Slices serve as effective units of work that eliminate subjective estimation problems inherent in story points
3. Fixed-cost pricing per slice removes bias and subjectivity from developer compensation
4. The life cycle of a slice includes development, integration, QA, and deployment phases with clear boundaries
5. Slices enable self-organizing teams where developers can barter and exchange work based on their strengths