AI’s Midas problem: When code is free, features are worthless

Every organization that adopted AI-powered coding tools such as Claude Code, Cursor, or Copilot faces a common issue: when output becomes free, relying on volume alone no longer proves successful. In fact, accelerating software development with AI might be eroding the very elements that make software valuable. It's time to rethink what truly adds value to software and why aggressively producing an abundance of code and features leads to failure.

The Midas Touch Ovid's story tells of King Midas of Phrygia, who, after granting the satyr Silenus hospitality, was offered a wish. Midas chose that anything he touched would turn to gold. Testing his newfound power, he delighted in turning an oak twig and a stone into gold. However, his joy turned to despair when he realized that even his daughter's rose garden was ruined, turning the sweet-smelling blooms into gold.

This realization prompted him to ask for the removal of his gift. Similarly, leaders who have pushed AI into developers' work are now experiencing a parallel realization. While the transition is gradual, organizations and leaders are realizing that the return on investment may not be as promising as initially thought. Reports indicate that some companies are reducing budgets and curbing AI tool usage.

Microsoft and Meta have reportedly decreased internal spending on Anthropic's Claude, while Uber and Accenture are limiting employees' spending on AI tools. McKinsey's 2026 research suggests that despite record AI spending, earnings for 94% of enterprises have not improved, highlighting the economic challenges of this approach. The Economics of Plenty In Ovid's tale, Midas' gold quickly lost value due to its hyperinflationary effects, causing prices to plummet.

In the contemporary world, anything created using AI becomes a commodity. When you use AI to generate code or content, anyone else can replicate it using the same tools. Whether you're developing software or writing a book, the perceived value diminishes to the cost of the AI tokens consumed. Basic coding skills once held value as they translated human requirements into code.

However, with AI, this value is now negligible. The same applies to software that merely reflects user requirements translated into code. These outputs are considered worthless. Valuable software now needs to convey more than just the translation of user needs into code. The Modern Approach to Avoid AI's Ruin The solution lies in offering something beyond what AI can replicate.

Just as Midas had to wash in the Pactolus River to rid himself of his golden touch, software developers must find ways to distinguish their work from AI-generated outputs. This requires moving away from the feature-factory model, where the primary value is in translating user wants into code. Instead, teams should focus on deep understanding of user jobs and challenges, enabling them to generate innovative ideas and insights that users wouldn't discover independently.

This approach, as highlighted by DORA's State of DevOps report, emphasizes user-centricity and can lead to improved product performance, regardless of delivery speed. The key is not to simply replicate user requests but to comprehend the underlying work and challenges, crafting software that provides genuine value. Embracing this user-centric approach means delivering something unique that addresses the core needs of users, going beyond the superficial features that AI can easily replicate.