Quoting Andrej Karpathy
As AI makes software creation nearly effortless, Andrej Karpathy observes that his personal demand for software is growing exponentially, illustrating the Jevons paradox in tech.
- AI slashes software marginal cost, enabling instant custom single-use apps, visualizers, and test suites.
- Jevons paradox applies to software: greater efficiency fuels vastly higher demand.
- The developer's role shifts from writing code to defining requirements and orchestrating AI.
- Software will become like water—on tap—but we must rethink quality, ownership, and dependencies.
Just a few days ago, Andrej Karpathy — one of AI’s most respected voices — tweeted a brief but loaded observation after testing the latest AI coding tools: “I feel a lot of things changing as working software increasingly comes out on a tap. The Jevon's paradox kicks in and I feel my own demand for software growing substantially. You can ask for anything — explainers, visualizers, dashboards, bespoke single-use apps, a 10X test suite, auto-optimized code, giant research projects with custom HTML for the results, anything!”
That offhand remark captures a profound economic and technological inversion. When the cost of making software plummets toward zero, do we need less of it — or vastly more? Common sense suggests efficiency would reduce consumption. But in the 19th century, William Stanley Jevons noticed that more efficient steam engines led to more coal being burned, not less. The same paradox is now hitting software.
The barrier to software creation has collapsed
Previously, even a simple dashboard or script required programming skills and hours of work. Now, a natural-language description can yield a working version in minutes. Karpathy’s mention of “a full wandb that is hyper-specific just for your project” illustrates the shift: previously such a tool might take weeks to build; now it’s almost free. As a result, all those “not worth building” ideas suddenly become viable.
At scale, the effect is even more dramatic. Test suites can grow 10x without extra human effort, code optimization can be sprinkled everywhere, and research reports can generate polished HTML automatically. When friction disappears, consumption explodes. Every developer becomes a demand amplifier — constrained not by hands, but by imagination.
The developer’s role is being rewritten
The impact goes beyond productivity. The core skill shifts from writing code to defining good problems and evaluating AI-generated outputs. Karpathy’s “anything door” means we can get results instantly, but we need the judgment to specify what we want and stitch the pieces into reliable systems.
Quality assurance will also transform. When tests multiply automatically and code review is partly delegated to AI, we need new trust mechanisms. How do you verify the AI’s statistics? How do you ensure auto-optimized code doesn’t introduce security bugs? These become the new engineering challenges.
The dark side of “software on tap”
Karpathy’s water-tap metaphor is vivid, but piped water also reminds us that infrastructure is centralized and quality is controlled by a few. When most software flows from a handful of AI platforms, dependency becomes deep. The explosion of bespoke, single-use tools could create a maintenance nightmare — unofficial apps that nobody owns yet everyone relies on.
That said, this is a thrilling inflection point. Karpathy’s insight shows we’re on the brink of software abundance. A developer’s worth may soon be measured not by lines of code, but by “how many unique problems were solved.” For anyone who works with software, it’s time to sharpen your sense of demand — because the ability to build is no longer the bottleneck. Deciding what to build is the new superpower.
Analysis by BitByAI · Read original