Quoting James Shore

2026-05-11 · Source: Simon Willison's Weblog · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning · Depth: Fundamental Awareness, quick

Summary

AI coding agents must significantly reduce software maintenance costs to provide a net benefit, according to James Shore. The author argues that if an AI agent doubles coding output, it must simultaneously halve maintenance costs to avoid a net increase in overall expenses. Failing to achieve this inverse relationship means that increased coding speed merely trades temporary productivity gains for a permanent increase in maintenance burden. The mathematical model presented illustrates that doubling output while maintaining current maintenance costs effectively doubles total maintenance expenses, emphasizing the critical need for AI to actively decrease, not just stabilize, these costs.

Key takeaway

For engineering leaders evaluating AI coding agents, your primary metric should be the agent's ability to reduce maintenance costs, not just increase code generation speed. Ensure that any productivity gains are directly correlated with a proportional decrease in long-term maintenance overhead, or you risk escalating total operational expenses significantly. Prioritize tools demonstrating clear, measurable reductions in code complexity and bug rates.

Key insights

AI coding agents must inversely reduce maintenance costs to justify increased code output.

Principles

• Productivity gains must offset maintenance costs.
• Increased output without cost reduction is detrimental.

In practice

• Evaluate AI tools based on maintenance cost reduction.
• Prioritize AI agents that simplify existing codebases.

Topics

• AI Coding Agents
• Software Maintenance Costs
• Developer Productivity
• Cost Reduction
• LLM Impact

Best for: CTO, VP of Engineering/Data, Executive, Software Engineer, Director of AI/ML

Related on AIssential

• Coders are refusing to work without AI — and that could come back to bite them — Widespread AI coding tool adoption masks dubious productivity gains and potential increases in maintenance costs.
• Measuring What Matters in the Age of AI Agents — Measure AI agent impact by treating them as team extensions, focusing on hybrid human-AI team performance.
• AI-coding agents are now the default. What comes next? — AI-coding agents are now default, demanding context optimization and human-in-the-loop processes for scalable productivity.
• The Real Cost of Agent-Written Software — The true costs of agent-written software extend beyond initial generation, encompassing debugging and reliability.
• Are AI agents actually slowing us down? — Over-reliance on AI coding tools prioritizes speed over quality, leading to increased tech debt and system outages.
• Advanced Claude Code Cost Tracking: How to Save 30% on Token Spend — Optimizing AI coding agent costs requires granular visibility and actionable configuration-level fixes, not just basic spend tracking.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.