At OpenAI DevDay 2026, Itamar Friedman, CEO of Qodo, explains how their platform integrates tribal engineering wisdom with AI models to enhance code review by capturing contextual insights and managing software evolution, thereby improving code quality and preserving critical developer knowledge. He also discusses the shift towards swarm engineering—collaborative AI workflows with multiple agents—and outlines Qodo’s future plans to manage complex development tasks holistically while emphasizing the continued importance of human oversight.
In this interview at OpenAI DevDay 2026, Danielle from the developer experience team introduces Itamar Friedman, CEO of Qodo, who explains that Qodo is designed to enhance the developer workflow by integrating tribal engineering knowledge with AI models like Codex. Qodo acts as a governance layer that not only reviews code but also maps software evolution over time, aiming to produce cleaner pull requests and better control over software projects. The company leverages OpenAI models to create a “wisdom-based” approach, combining raw code intelligence with the experiential knowledge that human developers accumulate over time.
Itamar distinguishes between intelligence and wisdom by emphasizing that while intelligence relates to raw data and information, wisdom is derived from experience and understanding of the entire software system. This wisdom is crucial for effective code review because it captures context and nuances that pure code analysis might miss. For example, Qodo helps identify downstream issues in code changes that appear correct locally but could cause significant problems in production, especially in complex environments like those involving database interactions.
The conversation also covers the evolution of AI engineering practices from prompt engineering to flow engineering and now swarm engineering. Itamar explains that while early AI work focused on simple question-answer prompts, modern approaches involve designing workflows where multiple AI agents (a swarm) collaborate with defined roles, goals, and guardrails. This swarm engineering approach is essential for building trustworthy AI systems that can handle complex software development tasks efficiently and reliably.
Regarding the code review process, Itamar highlights the importance of agreement between coding agents and review agents, with human intervention only when discrepancies arise. He stresses that continuous learning from mistakes is vital, with errors being codified into a “wisdom base” to prevent recurrence. This approach helps preserve tribal knowledge that might otherwise be lost when experienced developers leave an organization, ensuring that critical insights remain embedded in the software development lifecycle.
Looking ahead, Qodo plans to redefine what constitutes a “task” in software development, moving beyond individual pull requests to managing entire feature sets or capabilities end-to-end. They are developing tools like “work package triage” to help developers understand how multiple pull requests interconnect to complete larger tasks. Itamar encourages developers to experiment with OpenAI’s decision API and to design AI swarms tailored to specific workloads, optimizing for both effectiveness and cost-efficiency. The interview closes with a reflection on the importance of human oversight and training even as AI takes on more code review responsibilities.
Useful Links
