Spec-driven Development: How AI Changed Everything (And Nothing)
summary
Claims that requirements become the single source of truth and that AI can generate diagrams, domain models, code and tests from them; demonstrates a brownfield flow that extracts system use cases and ERD tables from existing code, Jira and Confluence, then regenerates a modern Spring Boot + Vaadin app using Cloud Code skills and MCP servers. Shows Playwright acceptance tests and Testcontainers-based browserless unit tests, reports ~26‑page use cases, ~3‑minute code generation runs, and warns humans remain responsible for review. Links code at github.com/simasch/aiup-petclinic and slides on SpeakerDeck.
AI-written summary · about
chapters (12)
- 0:00Spec-driven development intro
- 3:16Three pillars
- 5:23Spec vs model-driven
- 7:31AI Unify Process for greenfield
- 11:42Use cases for business apps
- 15:15Brownfield modernization
- 18:07Demo: Reverse engineering Spring PetClinic
- 27:24AI agent skills and MCP servers
- 32:24AI dev guardrails
- 34:35Demo: AI use case implementation
- 37:25Architecture impact on AI dev
- 40:15Conclusion and key takeaways
related talks
44mSpec-driven Development: How AI Changed Everything (And Nothing)
38mAI for Software Development: A Reality Check
59mAI is Eating the Developer Career! Wait, what? Maybe it's the other way around?
41mAccelerating Domain-Driven Design with AI
32mQuality engineering in the AI age: From testing AI to leveraging GenAI
26mAl and Software: The Good, The Bad, The Scary
30mIncorporating AI into your SDLC
47mThe Renaissance Developer: Building Better, Safer, & Quicker with AI-Powered Spec-Driven Development