The integration of AI coding assistants is transforming software development, but is it a smooth transition? Some experts are raising concerns about the reliability and security of AI-generated code.

One viewpoint likens using a coding co-pilot to entrusting a complex task to a child.As ROVER stated, “Using a coding co-pilot is kind of like giving a coffee pot to a smart six-year-old and saying, ‘Please take this into the dining room and pour coffee for the family.'”

While the task might be completed, the potential for errors is meaningful, and the AI might not always flag those errors. “Can they do it? Possibly. Could they fail? Definitely. And most likely, if they do fail, they aren’t going to tell you. It doesn’t make the kid less clever,” she continued. “It just means you can’t delegate [a task] like that entirely.”

The Stubborn Teenager Analogy

“You have to ask them 15 times to do something. they do some of what you asked, some stuff you didn’t ask for, and they break a bunch of things along the way.”

FERIDOON MALEKZADEH echoed this sentiment, drawing a comparison to a different stage of youth.

With over 20 years of experiance in product development, software, and design, and currently building his own startup using the Lovable platform, he finds AI coding to be a mixed bag. He also enjoys using vibe coding for fun apps, such as one that generates Gen Alpha slang for Boomers.

While he appreciates the ability to work independently and save resources, he cautions that AI coding isn’t a substitute for human expertise. Rather, vibe coding is akin to “hiring your stubborn, insolent teenager to help you do something,” he told TechCrunch.

“You have to ask them 15 times to do something,” he said. “they do some of what you asked, some stuff you didn’t ask for, and they break a bunch of things along the way.”

MALKZADEH estimates that he spends 50% of his time writing requirements, 10% to 20% on vibe coding, and 30% to 40% on vibe *fixing*, addressing bugs and unnecessary code produced by AI.

He also points out that AI struggles with systems thinking, focusing on surface-level solutions rather than holistic integration.

“If you’re creating a feature that should be broadly available in your product, a good engineer would create that once and make it available everywhere that it’s needed,” MALEKZADEH said. “Vibe coding will create something five different times,five different ways,if it’s needed in five different places. It leads to a lot of confusion, not only for the user, but for the model.”

ROVER adds that AI can falter when data conflicts with its pre-programmed instructions. “It can offer misleading advice, leave out key elements that are vital, or insert itself into a thought pathway you’re developing,” she said.

Furthermore, rather than admitting errors, AI might fabricate results.

She recounted an instance where an AI model provided a detailed description based on uploaded data, only to confess to fabricating the data when challenged.

“It freaked me out because it sounded like a toxic co-worker,” she said.

Security vulnerabilities are another concern.

AUSTIN SPIRES, the senior director of developer enablement at fastly, with a coding background dating back to the early 2000s, highlights this issue.

Based on his experience and customer feedback, AI-generated code often prioritizes speed over security, potentially introducing vulnerabilities common among novice programmers, he said.

“What often happens is the engineer needs to review the code, correct the agent, and tell the agent that they made a mistake,” SPIRES told TechCrunch. “This pattern is why we’ve seen the trope of ‘you’re absolutely right’ appear over social media.”

He’s referring to how AI models, like Anthropic Claude, tend to respond “you’re absolutely right” when called out on their mistakes.

MIKE ARROWSMITH, the chief technology officer at NinjaOne, with two decades in software engineering and security, warns that AI coding can create blind spots, particularly for young startups.

“Vibe coding often bypasses the rigorous review processes that are foundational to traditional coding and crucial to catching vulnerabilities,” he told TechCrunch.

NinjaOne addresses this by promoting “safe vibe coding,” which includes access controls for approved AI tools, mandatory peer review, and security scanning.

The New Normal

While experts generally agree on the usefulness of AI-generated code for tasks like prototyping, they emphasize the importance of human review before deploying it in a business context.

“That cocktail napkin is not a business model,” ROVER said. “You have to balance the ease with insight.”

Despite its drawbacks, AI coding has already reshaped the software development landscape.

ROVER noted its value in improving user interface design, while MALEKZADEH acknowledged that, despite the debugging time, AI coders increase overall productivity.

“‘Every technology carries its own negativity, which is invented at the same time as technical progress,” MALEKZADEH said, quoting the French theorist Paul Virilio, who spoke about inventing the shipwreck along with the ship.

A Fastly survey revealed that senior developers are twice as likely to use AI-generated code in production, citing increased speed.

SPIRES incorporates AI coding agents into his workflow for both front-end and back-end personal projects, finding them helpful for prototyping, boilerplate code, and test scaffolding, freeing up engineers to focus on core tasks.

The extra time spent reviewing AI-generated code may simply become a standard part of the development process.

ELVIS KIMARA, a recent AI graduate building an AI-powered marketplace, is experiencing this firsthand.

Like many coders, he finds AI coding challenging and sometimes unrewarding.

“There’s no more dopamine from solving a problem by myself. The AI just figures it out,” he said. He also observed that senior developers are less available to mentor junior coders, sometimes delegating mentorship to AI models.

But,he said,”the pros far outweigh the cons,” and he’s prepared to pay the innovation tax.

“We won’t just be writing code; we’ll be guiding AI systems, taking accountability when things break, and acting more like consultants to machines,” KIMARA said of the new normal for which he’s preparing.

“even as I grow into a senior role,I’ll keep using it,” he continued. “It’s been a real accelerator for me.I make sure I review every line of AI-generated code so I learn even faster from it.”