A new arXiv preprint from Microsoft researchers challenges the common practice of measuring AI's impact on software development purely by how much faster developers complete tasks. The authors argue this is like judging electricity by how much faster it spun a steam-era factory — it misses the transformative effect. Instead, they propose a fuller measure: what new work becomes possible when AI is available.

The study is based on a survey of 331 software developers at Microsoft. While the abstract does not reveal specific findings, it signals a shift in how AI's contribution might be evaluated. Rather than asking only "How much time was saved?" the researchers ask "What did developers do that they could not do before?"

This framing could have implications for how companies assess AI tools, productivity, and developer satisfaction. If adopted, it would move the conversation from efficiency gains to capability expansion — a more ambitious yardstick for AI's role in software engineering.