Axoniq Conference 2025 Day 2 | Event Modeling for Legacy Systems: Breaking the Greenfield Myth

🎯 Speaker:

• Martin Dilger - Consultant & Author of "Understanding Event Sourcing" (700 pages, 3,000+ readers)

• Based in Germany, specializing in event sourcing and event modeling adoption

💡 The Revolution:

First Truth Bomb: Event modeling isn't just for greenfield - it's a documentation and understanding tool for ANY legacy system

The Process:

• AI analyzes codebase → Generates event model → Understanding without code reading

• Event model → Code generation → Running application

• Works with CRUD, works with legacy, works with everything

📚 Book Background:

• Title: "Understanding Event Sourcing" (700 pages)

• Written in 3 months (1,000 words/day goal)

• 3,000+ readers

• Forwards by Adam (event modeling inventor) and Gabriel

💻 Live Demo Highlights:

Pet Registration Slice:

• Drew ugly UI in under 2 minutes (name, description, age fields + save button)

• Added "Register Pet" command

• Added "Pet Registered" event

• Defined information flow (name, description, age)

• Information completeness check (red arrow → green arrow)

• Exported JSON from Miro

• Fed to Axon Platform

• Generated working code

The 2,000 Character Hack:

• Platform limitation: 2,000 characters max

• Martin's solution: "Don't start generating before I tell you to start"

• Used Claude to trim JSON under limit

• Successfully bypassed restriction

• Generated complete flows, commands, events

🎬 Key Moments:

• "Drawing Ugly Screens is Part of the Process"

• Rule: Never more than 2 minutes per UI

• Philosophy: Speed over perfection in modeling phase

• Reality: Most people waste 10-15 minutes per screen

"Information Completeness Check"

• Red arrow = Missing information

• Green arrow = Complete information flow

• Prevents worst mistake: Wrong assumptions

• Real-time validation during modeling

"Slices Are My Unit of Work"

• Think in slices, work in slices

• Each slice = independently deployable functionality

• Start with one slice, implement it, ship it

• No need to complete entire model before coding

🎓 Key Lessons:

1. Biggest Problems Aren't Technical

• Technology is the easy 20%

• Real challenge: Making everyone agree on the problem

• Event modeling facilitates alignment

2. Nobody Has Greenfield Projects

• Everyone deals with legacy systems

• Event modeling works beautifully with existing code

• Use it to understand, then decide: extract or document

3. Event Modeling → Any Stack

• Could be event sourcing

• Could be SQL inserts

• Could be anything

• It's about understanding, not implementation

4. AI as Event Modeling Accelerator

• Extract models from code automatically

• Generate models from requirements

• Starting point for team discussions

• Human discussion remains the critical part