In the age of vibe coding, trust is the real bottleneck

AI can now write code faster than a human can possibly type. With “vibe coding” tools like Anthropic’s Claude Code and OpenAI’s Codex, developers are gleefully building—and shipping—at a pace that would have been unthinkable just a year ago. Even Claude Code’s creator, Boris Cherny, has boasted that the latest version was written entirely by—yes—Claude Code.

But while vibe coding may be fast, it can also introduce subtle bugs and vulnerabilities. And human error hasn’t gone away: Claude Code is now under scrutiny after its own source code was accidentally leaked this week due to a packaging mistake.

For enterprises, these kind of vulnerabilities are a nonstarter. At large companies with sprawling codebases, it’s not just about writing code faster—it’s about ensuring that code is correct, secure, and compliant with internal systems and external obligations. As AI tools begin to generate production-ready code automatically, the bottleneck is shifting from writing software to verifying it. And at enterprise scale, where millions of code changes can flow through a system each year, even small errors can quickly compound into major risks.

That got me thinking about an interview I did two years ago with Itamar Friedman, cofounder and CEO of Qodo, an AI code review tool that has just raised $70 million to tackle what he calls the growing problem of “AI slop” in codebases.

When I first spoke to Friedman in early 2024, when the company was called CodiumAI, he talked about “flow engineering”—a system where one model generates code and another critiques it, adding layers of testing and reflection. But even then, it was clear that generating code was considerably easier than making sure it is accurate and works well, and that “code integrity” was key. 

In a chat with Friedman yesterday, he argued that today’s AI coding tools, powered by LLMs, are designed to complete tasks, not to question them—making a separate “governance and trust layer” essential to determine what should (and shouldn’t) ship.

“AI is not enough when you’re talking about real-world software quality and code governance,” he said. “What you need, actually, is official wisdom.” He explained that as a developer in a big organization, creating quality code isn’t just about being smart. It’s about knowing how a specific company does things—all the tribal knowledge within the organization. 

Qodo, he explained, analyzes how developers in an organization actually write and review code—looking at pull requests, comments, and past changes—and turns that into a set of rules that define what “good” looks like for that company. Those rules are then enforced automatically, flagging new code that violates them.

In the age of AI, the challenge for enterprises is that they want to move faster, but don’t have the freedom to change their codebases unless they can be sure that code will remain trustworthy. 

“That’s the gap we’re trying to close,” said Friedman, who spent three years as a director of machine vision at Alibaba before launching what is now Qodo in 2022, just a few months before ChatGPT launched. Qodo clients, including Walmart, Nvidia, Ford and Texas Instruments, want to move fast, he explained, but they also know their systems depend on layers of accumulated knowledge and constraints. 

Today’s vibe coding landscape, he added, overestimates how much these tools can be trusted in the short term—and underestimates how much a trust layer is needed to make them viable in the real world for the long haul.

With that, here’s more AI news.

Sharon Goldman
sharon.goldman@fortune.com
@sharongoldman

FORTUNE ON AI

Asia’s AI playbook gets a reality check as the Iran war sends energy prices higher and snarls supply chains – by Angelica Ang

AI ‘slop’ is flooding YouTube Kids—and more than 200 groups and experts are calling for a ban – Catherina Gioino

AI models will secretly scheme to protect other AI models from being shut down, researchers find – by Jeremy Kahn

Anthropic mistakenly leaks its own AI coding tool’s source code, just days after accidentally revealing an upcoming model known as Mythos – by Beatrice Nolan 

AI IN THE NEWS

Microsoft CFO’s AI spending runs up against tech bubble fears. This terrific new Bloomberg profile details how Microsoft CFO Amy Hood has emerged as one of the most powerful—and controversial—figures shaping the company’s AI strategy, tasked with threading the needle between runaway infrastructure spending and the risk of falling behind in the AI race. According to Bloomberg, Hood made the call in late 2024 to pause parts of Microsoft’s massive data center buildout, questioning overly optimistic demand forecasts—a decision that rattled investors and may have contributed to today’s capacity shortages as AI demand surged beyond expectations. Known internally for her intense scrutiny and cost discipline, Hood has helped keep Microsoft’s margins stable even as rivals open the spending floodgates, but her cautious approach now sits at the center of a high-stakes dilemma facing all Big Tech: how to invest aggressively enough to win in AI without overshooting in what remains an uncertain—and potentially bubble-like—market.

How AI helped one man (and his brother) build a $1.8 billion company. According to the New York Times, entrepreneur Matthew Gallagher used a suite of AI tools to launch Medvi, a telehealth company selling weight-loss drugs, with just $20,000 and essentially no staff—scaling it to $401 million in revenue in its first year and projecting $1.8 billion this year with only one full-time employee (his brother). By relying on AI for everything from coding and marketing to customer service and analytics—while outsourcing regulated functions like doctors and fulfillment—Gallagher dramatically compressed the traditional startup playbook, illustrating how AI is enabling ultra-lean, hyper-scalable companies. There are also some glitches and hallucinations, and the need for some human oversight remains. Still, the broader trend threatens to reshape hiring, productivity, and the very idea of what a “company” looks like in the AI era.

Apple kicks vibe coding app out of App Store, escalating crackdown. Apple has escalated its crackdown on “vibe coding” apps by removing the AI-powered app builder Anything from the App Store, the Information reported. The company cited rules against apps executing unreviewed code. The move follows earlier efforts to block updates to similar tools, which let non-developers create and modify apps using AI, and reflects Apple’s growing concern that such platforms could flood the App Store with low-quality or dynamically changing software that bypasses its review process. While Apple says it’s simply enforcing existing guidelines, the crackdown also raises competitive and regulatory questions, especially as vibe coding tools gain traction and begin to challenge traditional development workflows—including Apple’s own Xcode ecosystem.

EYE ON AI NUMBERS

53%

That's how many US businesses would allow AI agents to negotiate prices or terms directly with other AI agents on their behalf, according to Visa's new Business-to-AI (B2AI) Report, conducted in conjunction with Morning Consult.  

The report highlighted how AI is already influencing demand overall: Nearly 40% of Americans have made a purchase they normally would not have considered as a result of using an AI agent or tool, which the report said is an early indication that intelligent systems are beginning to shape how people discover and decide what to buy.

Other notable stats: The survey found that 71% of businesses say they are willing to optimize products, offers and experiences specifically for AI agents, while 77% are already using or piloting AI in their operations.

AI CALENDAR

April 6-9: HumanX, San Francisco. 

June 8-10: Fortune Brainstorm Tech, Aspen, Colorado. Apply to attend here.

July 6-11: International Conference on Machine Learning (ICML), Seoul, South Korea.

July 7-10: AI for Good Summit, Geneva, Switzerland.

August 4-6: Ai4, Las Vegas, Nevada