Cursor Developer Habits Report (1-min read)

According to the 2026 Spring Developer Habits Report based on Cursor data, the software development landscape is rapidly changing due to the emergence of AI agents.

Development speed has doubled year-over-year, and the amount of code added per PR (Pull Request) has also increased by about 2.5x. In particular, Mega PRs of 1,000+ lines are becoming common, and the number of Tool calls by agents has increased by an average of 30%, indicating rising task complexity. In addition, the survival rate of AI-generated code (the rate at which it is still retained after 60 minutes) has risen to 81%.

Request costs vary by up to 9x depending on the model, but the cost per accepted line of code narrows this gap to about 7x. This shows that high-performance models are compensating for cost efficiency by accurately generating more code.

AI usage is heavily concentrated among a small number of power users. The Gini score is above 0.7, with the top 1% of developers accounting for a significant share of AI-generated lines, cost spending, and token consumption.

The surge in input tokens and Cache-read tokens signals that agents have gained a kind of 'working memory' that allows them to perform more complex tasks. Coding agents are now evolving beyond simple tools into agentic systems that automatically build and maintain software.