Fortune's AI reporter Sharon Goldman has published an analysis arguing that the biggest challenge in AI-assisted development is no longer writing code — it is verifying it. With 77% of businesses already using or piloting AI in operations and tools like Claude Code capable of generating entire applications from natural language descriptions, the bottleneck has shifted from production to trust. As Qodo CEO Itamar Friedman puts it: 'AI is not enough when talking about real-world software quality.'
Qodo, an AI code review startup working with clients including Walmart, Nvidia, Ford, and Texas Instruments, has raised $70 million in Series B funding to build what it calls the governance layer for AI-generated code. The company's approach analyses organisational code patterns to establish automated rules — what Friedman describes as 'official wisdom' — ensuring that AI-generated code meets internal standards, security requirements, and compliance policies before it reaches production.
The timing is pointed. Fortune notes that Claude Code's source code was leaked due to a packaging error just days earlier, and that Boris Cherny — who created Claude Code — had boasted that the tool's latest version was entirely written by Claude Code itself. The article highlights a growing paradox: AI tools are powerful enough to write all the code, but enterprises cannot deploy that code without human or automated verification at scale. With millions of code changes flowing through large organisations annually, manual review is no longer feasible.
For context engineers, this story captures a fundamental shift in what it means to be a developer in 2026. The value is moving from code generation to code governance — understanding what AI produces, ensuring it meets quality standards, and building the verification infrastructure that lets organisations trust AI-generated software at scale. Teams that master this verification layer will have a significant competitive advantage.
