One-sentence headline summary

Economic incentives will likely force AI models to prioritize high-quality, maintainable code over "slop" as businesses seek to reduce the long-term costs of software development and maintenance.

Key points

• Greptile’s 2025 State of AI Coding report shows lines of code per developer increased from 4,450 to 7,839 as AI tools became standard.
• Median pull request sizes grew by 33% between March and November 2025, contributing to concerns regarding software brittleness and increased system outages.
• Industry experts like Andrej Karpathy and Node.js creator Ryan Dahl have noted that AI agents often produce bloated, messy, or repetitive code.
• High-quality code is defined by simplicity and ease of modification, which reduces the token usage and compute power required for AI-driven development.
• Market competition among AI model labs will eventually favor tools that produce efficient, scalable code to minimize operational expenses for companies.

The current surge in AI-generated "slop" threatens to increase technical debt and software instability across the industry. However, as AI adoption matures, the economic necessity of reducing compute costs will likely shift development standards toward cleaner, more efficient code generation.