Microsoft CEO Nadella warns against wasteful AI token use but admits developers will soon oversee AI agents instead of writing code themselves.

What happened

Microsoft CEO Satya Nadella cautioned against "token-maxing"—using the most powerful AI models for routine tasks—arguing that the productivity gain must justify the token cost. He also outlined a vision where developers will no longer write code but instead oversee hundreds or thousands of AI agents, a skill he calls "cognitive coverage."

Why it matters

Nadella's warning suggests that indiscriminate use of expensive AI models wastes resources and won't drive real economic growth. At the same time, his vision for agent-driven development signals a fundamental shift in how software work will be structured, requiring developers to understand code they did not write rather than produce it themselves.

What to watch

Nadella acknowledged his own token-maxing habit, admitting "it is addictive," which underscores the tension between the efficiency argument and the real appeal of powerful AI systems to users and organizations alike.