Breaking conventions for more performant models
summary
Argues that intentionally breaking domain conventions simplifies models and boosts performance, with concrete patterns: treat day boundaries as business-specific (metro end-of-service ~2–3am, trading day flips at 16:00), model booking parties as modeling with roles (guest, company, hotel) to avoid wrong assumptions, chunk batch jobs by domain units (weeks/regions) to balance work, and convert continuous calendars to bitsets for fast availability checks using bitwise ops (one CPU cycle). Notes LLMs give better designs when prompted with bounded context.
AI-written summary · about
chapters (13)
- 0:00The dangers of realism
- 2:51Modeling as compression
- 5:14Rules for effective models
- 7:36Leveraging AI for modeling
- 8:53Breaking conventions in modeling
- 14:43Applying bounded contexts
- 17:00Modeling with roles
- 21:20Different things, same models
- 23:26Optimizing search and indexing
- 25:55Domain-driven batch processing
- 29:07Shifting to parallel universes
- 33:24Implementing speculative loading
- 38:00Conclusion
related talks
46mModel Mitosis: Stop making the wrong choice between microservices and monolith
50mModel Mitosis: a dynamic pattern to deal with model tensions
51mDesigning for Performance
42mHigh-Speed DDD (revisited)
48mModel Mitosis: Stop Confusing Microservices and the Monolith
50mAll Models Are Wrong: Lessons from Real DDD Projects
47mThe Joy of Multiple Models
43mArchitecture antipatterns and how to beat them 1/2