Developers Refuse to Work Without AI, Prompting New Questions on Productivity and Costs

A METR research team discovered that most developers now balk at performing any coding task without artificial‑intelligence assistance. The lab’s follow‑up survey found programmers believe AI doubles their value, yet recent reports from Amazon, Uber and independent researchers suggest the promised productivity boost may be illusory, with token‑tracking systems shut down and maintenance problems on the rise. The clash between developer reliance on AI and mounting expenses is reshaping how firms evaluate coding tools. Read more

May 30, 2026

AI Coding Tools Fail to Boost Productivity, Companies Waste Millions

A new study by METR could not replicate a 2025 experiment that claimed AI made developers more productive, as participants refused to work without AI assistance. Subsequent surveys and corporate reports reveal that AI coding tools often slow development, generate more bugs, and inflate costs. Amazon halted a token‑tracking leaderboard after employees gamed it, while Uber burned through its entire AI budget in four months without measurable gains. Experts warn that the industry’s reliance on AI-driven token consumption as a productivity metric is misguided and call for stronger code‑review practices. Read more