Specification-driven development and model-driven architecture with AI
summary
Argues that large language models finally solve the old MDA bottleneck by acting as the code generator, while the new problem is nondeterministic output that requires verification; proposes using BDD specs as tests and iterating LLM prompts until tests pass, turning specs into executable acceptance checks. Claims microservices benefit far more than monoliths because small, replaceable units are generatable and cheap to regenerate, and cites five years of experience validating that smaller services are reproducible and easier to verify.
AI-written summary · about
chapters (8)
- 0:00Welcome and introduction
- 0:54Model-driven architecture basics
- 2:17AI-driven development shift
- 4:30Why generators failed before
- 5:31Verification challenges with AI
- 6:14Behavior-driven design with AI
- 7:19Microservices from AI models
- 9:07Benefits of smaller units
related talks
05mModel Driven Software: Putting the Cart Before the Egg
47mTDD & Generative AI: A Perfect Pairing
44mSpec-driven Development: How AI Changed Everything (And Nothing)
30mUsing AI to Create Better Requirements and Designs
35mRewilding Software Engineering - AI, Tools and Human Decisions
41mAccelerating Domain-Driven Design with AI
23mStop Making LLMs Deterministic: A Spec-Driven Software Factory
46mTDD & generative AI - a perfect pairing?