Here is the dirty secret of the vibe coding revolution: speed was never the hard part. Trust is.

Bloomberg just ran a weekend feature on "AI FOMO," the growing anxiety among professionals watching colleagues ship apps overnight using tools like Claude Code and OpenAI Codex. Fortune followed with a deeper question: if AI can write code faster than any human, who checks whether that code actually works?

The answer, increasingly, is nobody.

Itamar Friedman, CEO of code review startup Qodo, which just raised $70 million, calls it the "AI slop" problem. Developers are shipping code at unprecedented speed. But AI models are designed to complete tasks, not to question them. They do not ask whether the code violates internal security policies. They do not know that your company handles healthcare data differently from ad data. They just ship.

The irony is almost poetic. Anthropic, which makes Claude Code, accidentally leaked its own source code this week because of a packaging mistake. The tool built to write perfect code could not keep its own codebase secure. That is not a bug. That is a warning.

For enterprises like Walmart, Nvidia, Ford, and Texas Instruments, all Qodo clients, the calculus is clear. They want the speed. They cannot afford the risk. Millions of code changes flow through these companies every year. A single vulnerability in a vibe-coded module could cascade through production systems.

Friedman argues the industry needs a "governance and trust layer" that sits between AI code generation and production deployment. His company analyzes how developers at an organization actually write and review code, then enforces those standards automatically on AI-generated output. Think of it as a corporate immune system for vibe-coded software.

The bigger picture is this: we have massively overestimated how much we can trust AI coding tools in the short term, and massively underestimated how much infrastructure we need to make them safe for the long term.

Vibe coding is not going away. It is too fast, too cheap, too democratizing. But the era where "it works on my machine" is good enough? That is already over. The companies that figure out the trust layer first will own the next decade of software. The ones that do not will wish they had.

Coverage sourced from Bloomberg, Fortune, and Qodo press materials.