You Need AI That Reduces Maintenance Costs

James Shore’s blog post “You Need AI That Reduces Maintenance Costs” argues that the benefits of AI coding agents are only realized if they also lower the cost of maintaining the code they produce. He explains that if an LLM doubles your output, you must halve your maintenance cost to keep total cost unchanged; otherwise, the cost quadruples. Shore warns that a temporary speed boost can become a permanent financial burden if maintenance costs do not fall in proportion. The article stresses that the math only works when the LLM actually decreases maintenance, not just keeps it steady.

He illustrates the trade‑off with a simple two‑by‑two example: double output and double cost equals quadruple cost; double output and steady cost equals double cost. Shore calls this “permanent indenture” if the cost increase is not offset. The post urges developers to measure maintenance per line of code and to evaluate whether the productivity gains justify the added expense. He concludes that without a clear maintenance‑cost reduction, relying on AI agents can backfire.