Vibe coding has turned senior devs into ‘AI babysitters,’ but they say it’s worth it

Carla Rover once spent 30 minutes on sob after he had to restart a project that she code.

Rover has been in the industry for 15 years and mainly works as a web developer. She is now building a startup, in addition to her son, who creates adapted machine learning models for market places.

She called an atmosphere that codes a beautiful, endless cocktail napkin on which you can constantly sketch ideas. But dealing with code generated by AI that people hope to use in production can be ‘worse than babysitting’, she said, because these AI models can ruin the work in ways that are difficult to predict.

She had encrypted herself to AI in a need for speed with her startup, just like the promise of AI tools.

“Because I had to be quick and impressive, I took a shortcut and don’t scans those files after the automated assessment,” she said. “When I did it manually, I found so much wrong. When I used a third -party tool, I found more. And I learned my lesson.”

She and her son start their entire project again – hence the tears. “I gave it up as the Copilot was an employee,” she said. “It isn’t.”

Rover is like many experienced programmers who turn to AI for coding help. But such programmers notice that they also behave like AI-Babysittershet rewrite and facts of the code that the AI ​​spits out.

A recent report from Fastly from Content Delivery Platform Company discovered that at least 95% Of the nearly 800 developers surveyed, they spent extra time on repairing Code generated by AI, where the load on such verification is the most serious on the shoulders of senior developers.

These experienced coders have discovered problems with code generated by AI, ranging from hallucinating package names to removing important information and security risks. Excited, not checked, an AI code can leave a product much more buggy than what people would produce.

Working with code generated by AI has become such a problem that it has given a new task of the company code that is known as “Vibe code cleanup specialist. ”

WAN spoke with experienced codingers about their time with the help of AI-generated code about what they see as the future of the coding of the atmosphere. Thoughts varied, but one thing remained for sure: the technology still has a long way to go.

“The use of a coding co-pilot is a bit like giving a coffee pot to a smart six-year-old and say:” Please take this to the dining room and pour coffee for the family, “said Rover.

Can they do it? Possible. Can they fail? Certainly. And most likely, if they fail, they won’t tell you. “It doesn’t make the child any less smart,” she continued. “It just means you can’t delegate [a task] Like that all. ‘

“You are absolutely right!”

Feridoon Malekzadeh also compared the atmosphere coding with a child.

He has worked in the industry for more than 20 years and held various roles in product development, software and design. He builds his own startup and uses a heavy atmospheric coding platform Lovable, he said. For fun, he also codes apps such as one that generates Alpha Snake for Boomers.

He likes it that he is able to only work on projects, save time and money, but agrees that atmosphere coding is not like hiring a trainee or a junior coder. Instead, coding the atmosphere is related to “hiring your stubborn, indiscriminate teenager to help you do something,” he said WAN.

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

Malekzadeh estimates that he spends about 50% of his time writing requirements, 10% to 20% of his time on mood coding, and 30% to 40% of his time on atmosphere determination -The remedy of the bugs and the “unnecessary script” made by ai-written code.

He also does not think that coding atmosphere is best in thinking of systems – the process of seeing how a complex problem can influence a general result. AI-generated code, he said, tries to solve more problems at surface level.

“If you make a function that should be available in general in your product, a good engineer would make it once and make it available wherever it is needed,” said Malekzadeh. “Vibe coding will create something five different times, five different ways, if necessary in five different places. It leads to a lot of confusion, not only for the user, but also for the model.”

In the meantime, Rover notes that AI “against a wall” when data conflicts with what it was difficult to do. “It can offer misleading advice, omit important elements that are of vital importance or insert themselves in a line of thought that you are developing,” she said.

She also discovered that instead of giving up mistakes, the results will produce.

She shared another example with WAN, where she questioned the results that an AI model initially gave her. The model began to give a detailed explanation and pretended it used the data she uploaded. Only when she called it does the AI ​​model admit.

“It drove me crazy because it sounded like a poisonous colleague,” she said.

Moreover, there are security problems.

Austin Spires has been the senior director of developer Enablement at Fastly and Coding since the early 2000s.

Through his own experience – together with chatting with customers – he likes to build atmosphere code what is fast instead of what is ‘good’. This can introduce vulnerabilities to the code of the species that very new programmers tend to make, he said.

“What often happens is that the engineer has to assess the code, corrects the agent and the agent tells that they have made a mistake,” Spiers told WAN. “This pattern is why we have seen the trope of ‘you are absolutely right’ on social media.”

He refers to how AI models, such as Anthropic Claude, tend to respond “You are absolutely right” when they are called for their mistakes.

Mike Arrowsmith, the Chief Technology Officer at the IT management software company Ninjaone, has been in software engineering and security for about 20 years. He said that the coding of the atmosphere creates a new generation of it and blind spots on the security places for which young startups in particular are susceptible.

“Vibe coding often bypasses the rigorous assessment processes that are fundamental to traditional coding and crucial for catching vulnerabilities,” he said WAN.

Ninjaone, he said, resisting this by encouraging ‘safe atmospheric coding’, with approved AI tools having access controls, together with compulsory peer review and, of course, security scanning.

The new normal

While almost everyone with whom we talked to the fact that AI generated code and atmospheric coding platforms are useful in many situations-such as mocking ideas, they all agree that human assessment is essential before building a company.

“That cocktaildap is not a business model,” said Rover. “You have to balance the convenience with insight.”

But despite all the mourning about his mistakes, atmosphere coding has changed the present and the future of the job.

Rover said that the coding of the atmosphere helped her enormously in making a better user interface. Malekzadeh simply said that, despite the time he spends, he still spent more done with AI coders than without them.

“Every technology has its own negativity, which at the same time is invented as technical progress,” said Malekzadeh, stating the French theoretician Paul Virilio, who spoke about inventing the shipwreck together with the ship.

The rapid research showed that senior developers were twice as likely to put AI into production generated by AI compared to junior developers, and said that technology helped them work faster.

Vibe coding is also part of Spires’ coding routine. He uses AI coding agents on various platforms for both front-end and back-end personal projects. He called the technology a mixed experience, but said it is good to help with prototyping, building a boilerplate or taking away a test; It removes subordinate tasks so that engineers can concentrate on building, shipping and scale products.

It seems that the extra hours spent on combing the atmosphere weed will simply be a tolerated load on the use of innovation.

Elvis Kimara, a young engineer, now teaches that. He has just graduated with a master in AI and builds an AI-driven marketplace.

Like many Coders, he said that the coding of the atmosphere has made his work more difficult and has often found an atmosphere that coded a joyless experience.

“There is no more dopamine of solving a problem alone. The AI ​​just discovers it,” he said. During one of his last jobs, he said that senior developers did not want to help so much to help young coding-bonds do not understand new atmospheric models, while others delegated mentoring tasks to the aforementioned AI models.

But he said, “The pros weigh much heavier than the disadvantages” and he is willing to pay the innovation tax.

“We will not only write code; we will guide AI systems, account for when breaking things and worn more like consultants on machines,” Kimara said about the new normal for which he is preparing.

“Even if I grow into a senior role, I will continue to use it,” he continued. “It’s a real accelerator for me. I make sure I look at every line of AI-generated code, so I learn from it even faster.”