Cursor's Spring 2026 report shows coding speed doubled year-over-year
Cursor's inaugural Developer Habits Report for Spring 2026 documents sweeping changes in agentic coding, including doubled year-over-year coding speed, larger PRs, deeper agent sessions, and a growing productivity gap between power users and everyone else.
Score breakdown
The report provides the first data-driven baseline from Cursor's platform showing that agentic coding has moved beyond individual acceleration into end-to-end automation of the software development lifecycle, with measurable productivity and cost-structure changes already visible in production data.
- 01Coding speed has doubled year-over-year according to Cursor data.
- 02Lines added per developer per week grew from ~3.6K in early 2025 to ~8.6K by mid-May 2026.
- 03Lines added per PR (p75) are up roughly 2.5x year-over-year, with the growth rate accelerating.
Cursor's inaugural Developer Habits Report for Spring 2026 presents data-driven findings across five themes. On developer acceleration, lines added per developer per week grew from roughly 3.6K in early 2025 to 8.6K by mid-May 2026, with growth visibly accelerating after the start of 2026. Lines added per PR (at the p75 level) are up approximately 2.5x year-over-year, and AI-generated code is surviving in codebases for longer periods, suggesting higher-quality output reaching production.
The report examines model economics from three angles — request cost, accepted-code efficiency, and the relationship between cost and benchmark performance — benchmarking seven model families on cost per line and cost per submit, revealing wide heterogeneity in unit economics. A key finding on context usage is that models are consuming more input tokens and cache-read tokens over time; because input tokens cost less than output tokens and cache-read tokens cost even less, this shift has favorable cost implications while also enabling agents to take on more complex tasks. The report also highlights a growing "power user gap," where the top 1% of developers are achieving significantly larger productivity gains than the broader user base, with the absolute gap widening as overall AI usage grows. Finally, the report frames the shift to automation as a new era in which AI coding tools are becoming infrastructure for automating the software development lifecycle end to end, rather than simply accelerating individual developers.
Key facts
- 01Coding speed has doubled year-over-year according to Cursor data.
- 02Lines added per developer per week grew from ~3.6K in early 2025 to ~8.6K by mid-May 2026.
- 03Lines added per PR (p75) are up roughly 2.5x year-over-year, with the growth rate accelerating.
- 04AI-generated code is surviving code review at higher rates than before.
- 05P99 power users are achieving significantly larger productivity gains than the rest of the user base, and the gap is growing.
- 06Models are shifting toward greater use of input and cache-read tokens, which cost less than output tokens.
- 07Seven model families were benchmarked on cost per line and cost per submit, revealing wide heterogeneity in unit economics.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 12, 2026 · 10:05 UTC. How this works →