Programming Deflation (tidyfirst.substack.com)

“Programming deflation” describes how AI-augmented coding is steadily driving down the time, cost and skill needed to write software. Economically this creates a tension: substitution suggests machines will replace coders, while Jevons’ paradox says cheaper code will unlock many more use cases and increase demand. Unlike destructive macro deflation, this is productivity-driven: lower costs and near-zero experimentation costs create a reinforcing loop—better tools speed creation of better tools—so we’re more likely to prototype and ship, producing a flood of mostly low-quality code alongside a smaller set of high-value, well-crafted systems. For the AI/ML community this means the battleground shifts from raw implementation to composition, evaluation and judgment. Technical implications: commoditized code amplifies the need for system design, API-driven composition, orchestration, observability, human-in-the-loop evaluation, and robust safety/quality metrics; model improvements will accelerate adoption and raise integration complexity. Practically, teams should use cheap tools to automate routine work, invest in integration and systems thinking, and cultivate “taste” — the ability to choose which problems deserve build effort. Those skills remain valuable whether the number of programmers shrinks or explodes, because scarcity moves from typing code to understanding what to build and how pieces fit together.