Sonar Summit 2026 | The self-healing applications decade is here

The rapid acceleration of artificial intelligence in software development has created an unexpected paradox. While generative AI tools have dramatically increased developer velocity—enabling pull requests to be created in minutes and code to ship faster than ever before—engineering teams are simultaneously struggling with increased code churn, higher issue volumes, and mounting technical debt. At Sonar Summit 2026, Boris Pascalv, CEO and co-founder of Logicar, explored this critical challenge alongside host Servanchy from Sonar's product marketing team. The conversation highlighted a fundamental blind spot in how the industry has approached AI-driven development: while optimizing for speed and productivity, organizations have neglected to optimize for decision quality.

The Velocity Paradox: Speed Without Direction

The statistics paint a troubling picture. Pull request volumes have more than doubled in some organizations every few months, yet this increase in velocity has come at a cost. Code duplication, stability issues, and review time have all increased significantly. Developers feel productive—and they are—but they're caught in an endless maintenance loop, fixing problems rather than building new features. The number of bugs per developer has actually increased substantially, despite tools becoming more capable. This paradox reveals a critical gap: organizations have become excellent at detecting problems through automated scanning tools like Sonar and observability platforms, but they lack a coherent decision-making framework to determine which problems actually warrant fixing.

The Signal-to-Noise Crisis

Engineering managers today are drowning in signals. Tools automatically detect code quality issues, security vulnerabilities, and production problems at scale. However, backlogs were never designed to capture and synthesize this overwhelming volume of evidence. Research cited during the session reveals that over 90% of production incidents had already generated signals somewhere in the system—either in a scanner, production monitoring system, or backlog—before they occurred. The problem isn't the absence of warnings; it's the inability to prioritize them effectively. Most engineering leaders aren't concerned about what's visible in their backlogs; they're worried about what's hidden within the signal noise they don't have time to examine.

Building a Decision Layer for Production Reliability

The solution, according to Pascalv, requires introducing an intelligent decision layer that helps managers prioritize work based on impact rather than volume. As AI continues to accelerate code generation and deployment, the bottleneck has shifted from engineering velocity to decision quality. Engineering managers must determine not just what could be fixed, but what actually should be fixed to maintain a sustainable balance between innovation and stability. This requires connecting signals from multiple sources—code quality scanners, observability systems, and production data—into a coherent framework that surfaces genuinely impactful issues while filtering out noise.

Key Takeaways

• The Velocity Paradox: AI acceleration of code generation has increased developer productivity while simultaneously increasing code quality issues, duplication, and technical debt accumulation
• Decision-Making is the Bottleneck: Engineering managers are no longer constrained by development speed but by the ability to prioritize which problems to address among overwhelming signal volume
• 90% of Incidents Were Predictable: Production incidents typically have pre-existing signals in scanners or backlogs, indicating the real problem is prioritization, not detection
• Signal-to-Noise Requires Intelligence: Modern development requires an automated decision layer that synthesizes evidence from multiple sources to identify high-impact issues worth fixing
• The Self-Healing Era Requires Balance: Sustainable development practices demand optimizing for decision quality alongside velocity to prevent technical debt spirals