Sonar Unveils AC/DC Framework to Cut AI Code Bloat After 41% Complexity Rise
Updated
Updated · InfoWorld · Jun 16
Sonar Unveils AC/DC Framework to Cut AI Code Bloat After 41% Complexity Rise
1 articles · Updated · InfoWorld · Jun 16
Summary
Sonar pitched its Agent Centric Development Cycle—guide, verify, solve—as a workflow to keep AI-generated code maintainable, reliable and secure before bloat compounds.
41% code-complexity growth and a 30% rise in static-analysis warnings persisted in 807 open-source projects using Cursor, Sonar said, citing Carnegie Mellon research showing early speed gains faded by month three.
4,400-plus Java benchmark tasks showed large model differences: GPT-5.4 High produced 1.159 million lines at an 81.05% pass rate, while Claude Opus 4.7 Thinking used 336,000 lines for an 82.52% pass rate.
200-line context limits, in-loop tests, static analysis and security scans are central to AC/DC, which Sonar argues scales better than human review as AI-heavy teams merge 98% more pull requests and review times climb 91%.
Sonar argues better models alone will not fix code decay; teams that build disciplined agent workflows now will be better positioned to ship enterprise-grade software over the next 18 months.
Is AI's coding speed creating a hidden 'technical debt bomb' for companies?
Can we teach AI to value elegant code, not just code that passes a test?
With AI automating junior tasks, are we erasing the training ground for future expert engineers?
From 84% Adoption to 29% Trust: Sonar’s AC/DC Framework and the Future of AI-Driven Code Quality
Overview
Sonar has introduced its Agent Centric Development Cycle (AC/DC) framework and open beta tools in response to the software industry's growing adoption of AI. As organizations shift their focus from just generating code to building robust workflows, there is a strong demand for AI solutions that not only identify but also solve problems, integrate smoothly into developer workflows, support governance, and reduce technical debt. The AC/DC framework is designed to meet these industry needs by enabling AI to resolve issues at scale without compromising trust, helping teams ensure that AI-generated code is trustworthy, secure, and maintainable.