Simon Willison: coding agents make minor refactors worth doing
Developer Simon Willison says cheap coding agents let him pursue small readability tweaks he'd otherwise skip.
Developer Simon Willison posted on X that coding agents are improving the quality of his code by making previously uneconomical refactors viable. Responding to Rhys Sullivan, Willison said that when he spots a tedious minor improvement — such as a readability tweak requiring edits across 20 or more places — he now delegates the work to a coding agent at effectively no cost to himself.
The remark speaks to a shift in developer economics as agentic coding tools mature. Small-scale refactors that humans routinely defer because the effort outweighs the payoff become worth doing when an agent handles the mechanical sweep. For teams building, buying, or using AI coding tools, the implication is that code quality gains may come less from agents writing new features and more from their willingness to execute the low-value, high-tedium cleanup work that human developers skip.