Seperation Of Scales
summary
Argues that design succeeds when abstractions pick a precise scale and stick to it, not when a topic is prettified; demonstrates that requests.get in two lines hides DNS TCP TLS and HTTP work and therefore is useful, shows tradeoffs: faster C/ASICs beat Python but cost development time, and complexity is conserved so abstractions only move it. Recommends starting from a smallest working system (Gall's law), think about information flow, prefer open standards like h5py and Matplotlib, and build replaceable interfaces.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 2:32Computers as layered systems
- 5:22Separation of concerns
- 10:59Scales in physics and software
- 17:28Constraints for freedom
- 24:59Trade-offs of abstractions
- 30:00Language and tool selection
- 34:01Building open source foundations
related talks
18mAPIs at Scale Through Minimalism in Code and Architecture
45mAbstraction Considered Harmful (and other opinions about writing maintainable code)
36mOn the Spectrum of Abstraction
27mScaling Your Front-End in a Realistic Way!
40m“Where We Can Go”
32mA Simpler Time: Balancing Simplicity and Complexity
26mBe Awesome By Being Boring
40mMore Code More Problems