Beyond the Known Knowns
summary
Argues that architecture is an executable organization of knowledge and that software development runs on incomplete, often invisible assumptions; uses Ariane 5 clamp bug (maiden launch fails ~38s after lift off, June 4 1996) to show how Ada code reuse and an unchecked 64→16-bit conversion causes catastrophe. Cites practical limits like the CAP theorem, cognitive biases (planning fallacy), the five orders of ignorance, and recommends iterative learning via the PDSA cycle, delaying commitment and treating architecture as testable hypotheses.
AI-written summary · about
chapters (10)
- 0:00Introduction and welcome
- 1:12Beyond the known knowns
- 7:58Limits of knowledge in software
- 18:09Known unknowns and planning fallacy
- 25:14Architecture as hypothesis
- 30:05Unknown unknowns and Ariane 5 failure
- 36:45Dependencies and assumptions in architecture
- 42:23Unknowable unknowns and roadmaps
- 46:54Design process and control
- 55:35Q&A and closing
related talks
55mArchitecture Blindspots, Biases, and Blunders: Avoiding Common Pitfalls in System Design
49mArchitecture as Knowledge
49mWhy software architects fail: and what to do about it
45mSoftware Architecture for Humans!
24mThe (too) Many Faces of Architecture
51mSoftware Architecture for Humans!
44mThe Philosophy of Architecture
08mLightning talks – 101 Things Software Never Learnt