Jevons Paradox: Why Cheaper Code Makes Worse Software
Cheaper generation doesn't mean the same code in less time. It means more code, more maintenance, and a self-reinforcing cycle with no natural brake.
Arthur Hertweck
Staff Engineer & Data Architect
I architect data and application layers for a platform serving millions of families. I've spent 15 years doing this work across healthcare, finance, and consumer tech, and the thing I keep coming back to is that the industry's hardest problems were never about code. They were about specification, comprehension, and judgment. AI just made that obvious. This site is where I map the territory and work through what to do about it.
Featured Series
15-Part Series
The Specification Age
AI didn't shift the bottleneck from generation to specification, the bottleneck was always specification. AI collapsed the pretense that generation was where the difficulty lived. This series examines what that means for engineering.
- 1 The Day Code Became Free
- 2 Three Constraints That Replaced One
- 3 Comprehension Debt
- 4 The Tautological Testing Trap
- 5 Jevons Paradox
- 6 When Every Metric Lies
- 7 The Expertise Pipeline Is Collapsing
- 8 Solution Monoculture
- 9 Your SDLC Was Designed for a World That No Longer Exists
- 10 The Trust Problem Has No Technical Solution
- 11 The Curation Thesis
- 12 New Roles and Skills for the Specification Age
- 13 A Measurement Framework
- 14 The Practitioner's Manifesto
- 15 Rebuilding the Onramp
Latest Writing
The Tautological Testing Trap
The oracle principle quietly breaks when AI writes both the code and the tests. Coverage rises, verification power falls, and the difference is invisible.
Comprehension Debt: The Liability Nobody Is Accounting For
Code that nobody designed produces systems that nobody understands. The maintenance costs are real and growing.