Stop writing software like it's 1999!
summary
Many teams still design for the 1990s so code is brittle and slow to change; the talk argues for feature-oriented folders and modular monolith structures, CI/CD and automated test pyramids from day one (pipelines pay back in weeks), avoiding premature microservices (one real example shows a team owning 27 services and suffering context and YAML chaos), and using event-sourcing/CQRS or specialized stores instead of heavy relational joins. A customer who tried “vibe coding” with LLMs fires devs and returns after four months because the generated system lacks engineering practices.
AI-written summary · about
chapters (10)
- 0:00Introduction and context
- 2:12Legacy of layered architecture
- 5:37Transition to feature-based code
- 14:22Microservices vs. monoliths
- 19:03Continuous delivery importance
- 26:01Automating infrastructure
- 32:55The role of automated testing
- 39:18Revisiting storage and database
- 45:30Architectural alternatives
- 50:23Engineering in the age of AI
related talks
50mTactics and strategies on software development: how to reach ...
57mBuilding Systems that are #neverdone
39mSoftware is a meal best served with different architectural spices
36mReinventing the wheel ... as a service
35mSoftware Engineering - Development in 100 Years Time
36mIt's Simple...
28mAvoid overengineering & big ball of mud- 7 lessons for maintaining the balance
57mForking the Future: Embracing Microservices Evolution