Jevons Paradox for Knowledge Work
A tutorial by Aaron Levie. Featured in the OTF curated resource library.
What Is Jevons Paradox?
In 1865, economist William Stanley Jevons observed something counterintuitive: as steam engines became more fuel-efficient, total coal consumption increased rather than decreased. More efficient engines made coal-powered work cheaper, which opened up entirely new applications that had been previously uneconomical.
This pattern — where increased efficiency leads to increased total consumption — has repeated throughout history. More efficient cars led to more driving. Cheaper data storage led to more data. Faster internet led to more internet usage.
The mechanism is simple: when something becomes cheaper, the demand for it expands faster than the per-unit cost decreases. The total amount consumed goes up, not down.
Applied to Coding
Now apply this to software development. AI coding agents are making code production dramatically cheaper — perhaps 5-10x in the near term. The naive prediction: developers will work less, and fewer developers will be needed.
Jevons Paradox predicts the opposite. When coding becomes 10x cheaper, organizations don't write 10x less code. They discover 100x more things worth coding.
Consider what's currently too expensive to build:
- Custom internal tools (companies use spreadsheets instead)
- Personalized user experiences (everyone gets the same UI)
- Comprehensive test suites (most code ships undertested)
- Documentation that stays current (most docs are stale)
- Accessibility features (retrofitted, if at all)
- Localization (limited to major languages)
All of these become economically viable when coding is 10x cheaper. The total volume of code written increases dramatically.
The Expanding Frontier of What's Worth Building
Hyper-Personalization
When building a custom UI variant costs almost nothing, every user gets a personalized experience. E-commerce sites, dashboards, and content platforms will generate per-user interfaces rather than one-size-fits-all designs.
Internal Tooling Explosion
Every department gets custom tools. Marketing gets a custom analytics dashboard. Sales gets a personalized CRM. Support gets specialized workflows. The internal tools market expands enormously.
Disposable Software
Software built for a single event, meeting, or project — then discarded. A custom app for a conference, a one-time data analysis tool, or a temporary landing page for a weekend promotion. Building becomes as cheap as writing a document.
Comprehensive Testing and Documentation
When generating tests and docs is nearly free, every project gets comprehensive coverage. Code quality increases across the entire industry as the economics of quality assurance shift.
New Roles and Opportunities
Jevons Paradox doesn't just increase volume — it creates new categories of work:
AI-Native Product Designers: People who think in terms of what AI can build, designing products that leverage AI capabilities from the ground up rather than retrofitting AI onto existing products.
Quality Orchestrators: With more code being generated, the role of ensuring quality, consistency, and security across AI-generated output becomes critical. This is a new specialization that didn't exist before.
Domain Translators: Experts in specific fields (healthcare, finance, education) who can translate domain knowledge into prompts and specifications that AI agents execute. They don't need to code — they need to specify clearly.
AI Workflow Architects: Engineers who design the pipelines and systems that enable autonomous AI coding at scale. They build the infrastructure that makes agent-driven development reliable and efficient.
The pattern from every previous technological shift: the total number of jobs in the ecosystem increases, even as individual tasks become automated. The new jobs require different skills.
Practical Implications
Invest in Specification Skills
The ability to clearly specify what should be built becomes the highest-leverage skill. This includes requirements writing, design thinking, and domain expertise — not just coding.
Build for Disposability
When building is cheap, maintaining is the expensive part. Design systems that are easy to rebuild rather than complex to maintain. Simplicity becomes even more valuable.
Embrace the Volume
Don't resist the increase in output. Build 3 prototypes instead of debating which approach is best. Test 10 landing pages instead of optimizing one. The economics of experimentation have fundamentally changed.
Focus on What AI Can't Do
Taste, judgment, user empathy, strategic thinking, and domain expertise remain human strengths. Double down on these skills — they become more valuable as implementation becomes commoditized.
