The Four Horsemen of Agentic Coding

Agentic coding is framed as a powerful but corrosive shift in software work, producing four interrelated harms: slop, alienation, deskilling, and team fallout. Slop describes a persistent, machine-produced code smell - functional but inscrutable patterns that models favor (examples include “word salad” prose and code-golfy, unreadable snippets) that make codebases hostile to human inspection and maintenance. That takeover of the code surface pushes people away from the repo into a design-at-a-distance role, eroding the hands-on craftsmanship that used to anchor software quality. Alienation follows: engineers lose kinesthetic connection to their tools and products, substitute agent-run workflows for direct coding, and consequently care less; features become half-baked, fixes become superficial, and personal pride in craft dwindles.

Deskilling amplifies the problem by changing incentives: an “easy button” reduces the need to learn core skills, shrinking pathways for early-career learning and threatening the longevity of current expertise while the claimed AI learning curve proves shallow. Team fallout completes the quartet - agents hollow out informal communication, mentorship, and the social rituals that surface diverse skills, turning teams into collections of agent operators rather than craftspeople. The net result is a productivity tradeoff that preserves short-term gains but damages curiosity, competence, and social cohesion with no clear remedy offered.