Slop Creep: The Great Enshittification of Software

7 hours ago

Slop creep refers to the gradual degradation of a codebase through small, seemingly reasonable decisions that collectively cause significant problems.
Coding agents can accelerate slop creep by making numerous small architectural mistakes quickly, compounding issues faster than human developers.
Agents lack holistic system understanding, often creating solutions that are correct in isolation but wrong for the overall system.
The absence of a natural 'circuit breaker' in agent-driven development allows problems to accumulate unchecked until they become severe.
Effective use of coding agents requires human oversight, particularly for one-way door decisions like data models and service boundaries.
Iterative refinement with agents, where humans review and correct initial drafts, produces better results than unsupervised agent output.
Planning phase is crucial when working with agents; vague instructions lead to slop, while detailed plans with code snippets yield cleaner results.
Poor code quality eventually affects user experience through subtle but persistent issues that degrade product quality.
The role of engineers is evolving from writing code to overseeing and directing agent output, ensuring system-wide coherence.
Current AI limitations mean human judgment remains essential for anticipating long-term consequences of architectural decisions.

Hasty Briefsbeta