42% of Code Comes From AI, Output Up 76%—But 96% of Developers Remain Skeptical
Surveys reveal that while 72% of developers use AI tools daily and AI‑assisted commits have risen to 42% with a 76% boost in code output, 96% still distrust AI‑generated code, only 0‑20% of tasks can be fully delegated, and 27% of the new output consists of work that previously didn’t exist.
Data
Sonar State of Code 2025 (1100 developers): 72% use AI tools daily, 42% of code commits are AI‑assisted, 96% doubt the reliability of AI‑generated code.
Greptile The State of AI Coding 2025 : individual output rose 76% (from 4,450 to 7,839 lines per month); medium‑size teams (6‑15 members) saw an 89% increase; pull‑request volume grew 33% while per‑file change density rose only 20%.
Anthropic 2026 Agentic Coding Trends Report : engineers use AI in ~60% of their work; tasks that can be fully delegated to AI are 0‑20%; ~27% of AI‑assisted output consists of tasks developers previously never attempted.
Three Observed Phenomena
More Use, Less Trust
AI now contributes to 42% of code, yet 96% of developers do not fully trust it. The core difficulty is judging the correctness of AI output after it is generated.
Output Rises Without Labor Savings
Code output increased 76% while fully delegable tasks remain 0‑20%, meaning developers do more work rather than less. Greptile data shows PR volume up 33% but per‑file change density up only 20%, indicating larger code changes without finer‑grained improvements and increased review pressure.
New Work Emerges
Approximately 27% of the additional output comprises tasks that were previously not undertaken—low‑priority, resource‑intensive, or otherwise avoided work—showing that AI reshapes what developers consider worth doing.
Full‑Process Cursor Open‑Source Project
The author ran a complete software development lifecycle using the Cursor tool, visualizing local AI Skills such as Codex, Claude code, and OpenClaw.
Key commands: npx you-skills – https://github.com/goodpostidea-tech/you_skills
npx skills add https://github.com/goodpostidea-tech/skills --skill you-skills– https://github.com/goodpostidea-tech/skills
During development, as code volume grew, the AI began to forget earlier‑established details (e.g., interface logic, configuration conventions), mirroring Greptile’s finding that larger PRs dilute the AI’s memory of prior context.
Changing Roles
Software development has progressed through abstraction layers (machine code → assembly → C → high‑level languages). Engineers now act as implementers —writing domain‑specific code quickly and accurately—and as orchestrators —evaluating cross‑domain solutions and verifying AI output.
Although 72% of developers use AI daily, only 0‑20% of tasks can be fully delegated; the remainder requires judgment, acceptance, correction, and re‑evaluation.
Open‑Source Project Resources
YouSkills: npx you-skills – https://github.com/goodpostidea-tech/you_skills
Skills repository:
npx skills add https://github.com/goodpostidea-tech/skills --skill you-skills– https://github.com/goodpostidea-tech/skills
Data Sources
Sonar · State of Code 2025
Greptile · The State of AI Coding 2025
Anthropic · 2026 Agentic Coding Trends Report
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
AI Engineer Programming
In the AI era, defining problems is often more important than solving them; here we explore AI's contradictions, boundaries, and possibilities.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.