← back
The State of AI Code Quality: Hype vs Reality — Itamar Friedman, Qodo
Takeaway
You don't break the AI productivity glass ceiling with better autocomplete — you break it with agentic, dynamically-learning quality workflows across the whole SDLC.
Summary
- Cites Qodo/Sonar/GitClear reports across thousands of devs and millions of PRs: 82-92% of devs use AI tools weekly/monthly, 60% of devs say >25% of code is AI-generated, 15% say >80%.
- Cursor/Copilot rules are only mostly followed, not rigorously — Claude Code's hyped security-review prompt actually says 'exclude denial of service', suggesting why cloud outages keep happening despite AI 'quality' tools.
- Code generation has a glass ceiling on productivity; the next leap comes from agentic quality workflows outside the IDE (review, testing, deployment) that learn dynamically.
- 67% of devs report serious quality concerns about AI-generated code; PR open rate up 97% while review time up 90% — reviewer is the new bottleneck.
code-qualityai-codingreview
Original description
AI is making code generation nearly effortless, but the critical question remains: can we trust AI-generated code for software that truly matters? Has it really become easier to build robust, high-quality systems? In this talk, we’ll separate hype from reality. Drawing on the State of AI Code Quality report, we’ll explore where AI tools excel, where they fall short, and which evaluation frameworks actually matter. We’ll unpack benchmarks, pitfalls, and lessons learned from deploying AI in real-world engineering workflows. Attendees will leave with a clearer understanding of how to measure, trust, and improve both AI-generated code and AI-driven code review—and what’s next for software quality in the age of AI. Speaker: Itamar Friedman | CEO, Qodo https://x.com/itamar_mar https://www.linkedin.com/in/itamarf/