Reverse Jevons Paradox

a day ago

Jevons paradox occurs when a decrease in resource cost leads to an increase in total usage, enabling new use cases that were previously unviable.
Examples include cheaper gas causing more driving and faster build systems increasing CI usage, both resulting in higher total spend.
The reverse Jevons paradox happens when increasing the cost of a resource reduces total usage, potentially to zero, eliminating beneficial activities like small code refactors.
This reverse effect represents a fundamental shift, not just a slowdown, and can be disastrous for codebase health when making changes overly difficult.

Drives tend to be nearly full regardless of size; free space is often only a few percent.
A poll showed about half of respondents had drives more than 75% full, indicating a common problem.
Increasing storage capacity by orders of magnitude does not solve persistent fullness.

AI-assisted coding increases software demand via Jevons paradox, similar to how cheaper computers led to more computing, not less.
Programmers face layoffs not due to AI replacing jobs, but due to shareholder-driven performative cuts to signal AI commitment, undermining software quality.
Software demand is elastic: lower development costs expand applications, from hobby projects (e.g., rabbit booking system) to commercial ventures (e.g., reimplementing Microsoft Word).

Dishwashers save little time because dish usage increases to match capacity, leading to more elaborate dish use.
AI similarly generates and fixes bugs, increasing code output without reducing work time, questioning efficiency gains.
Optimization often feels zero-sum; increasing productivity doesn't necessarily enhance well-being or happiness.

The incident involved intermittent failures in a search backend triggered by a slow dependency, not CPU throttling.
The false root cause (CPU throttling) was initially appealing but contradicted by topology, timing, and failure distribution evidence.
The actual cause was a combination of no timeouts on downstream calls and eager retries, creating a feedback loop that saturated the worker pool.

AI models have reached an inflection point where their output is worth more than the cost for real development work, leading to increased inference spending.
The author's annual AI inference spend grew from $100 in 2023 to $5000 in 2025, reflecting growing utility.
o1-pro was the first model consistently useful for professional coding, capable of debugging large files and saving hours per week.

Drives tend to be nearly full regardless of size; free space is often only a few percent.
A poll showed about half of respondents had drives more than 75% full, indicating a common problem.
Increasing storage capacity by orders of magnitude does not solve persistent fullness.

AI-assisted coding increases software demand via Jevons paradox, similar to how cheaper computers led to more computing, not less.
Programmers face layoffs not due to AI replacing jobs, but due to shareholder-driven performative cuts to signal AI commitment, undermining software quality.
Software demand is elastic: lower development costs expand applications, from hobby projects (e.g., rabbit booking system) to commercial ventures (e.g., reimplementing Microsoft Word).

Dishwashers save little time because dish usage increases to match capacity, leading to more elaborate dish use.
AI similarly generates and fixes bugs, increasing code output without reducing work time, questioning efficiency gains.
Optimization often feels zero-sum; increasing productivity doesn't necessarily enhance well-being or happiness.

The incident involved intermittent failures in a search backend triggered by a slow dependency, not CPU throttling.
The false root cause (CPU throttling) was initially appealing but contradicted by topology, timing, and failure distribution evidence.
The actual cause was a combination of no timeouts on downstream calls and eager retries, creating a feedback loop that saturated the worker pool.

AI models have reached an inflection point where their output is worth more than the cost for real development work, leading to increased inference spending.
The author's annual AI inference spend grew from $100 in 2023 to $5000 in 2025, reflecting growing utility.
o1-pro was the first model consistently useful for professional coding, capable of debugging large files and saving hours per week.

Hasty Briefsbeta