'Vibe coding’ and other ways AI is changing who can build apps and how
Doher Drizzle Pablo was drowning in travel receipts. After her company transferred her to Sweden from the Philippines last year, she’d started visiting clients in at least two countries a month, and the paperwork was so tedious it was starting to cost her money.
“I was super frustrated because my husband kept telling me I wasn’t filing my reimbursements,” Pablo recalls. She laughs now, but at the time, she was fed up and overwhelmed by the spreadsheets, so she decided to tackle the chore using AI.
She didn’t know it yet, but she was about to try something new: Building software by talking through an idea with an AI tool, which writes the code to make it real.
“Vibe coding,” a term coined by prominent computer scientist Andrej Karpathy, grabbed attention earlier this year as one example of this trend. Now the broader approach of using AI to turn conversation into creation is reshaping how software gets built — opening the door to newcomers and freeing developers to move faster and experiment.
Building by talking: the promise of conversational tools
Pablo isn’t a coder and was glad she didn’t need to learn any programming languages that day. She simply began chatting about her situation in plain language with an AI-driven feature called Plan Designer in Microsoft Power Apps. Within two hours, she was using her own custom expense management app and began amazing her manager as she filed receipts on time or even early, she says.
Pablo, who grew up in a rural area and didn’t even have a computer until she moved to Manila as an adult, now uses AI tools, including Copilot within Power Platform, to create programs that help with everything from collating different calendars to event planning to sharing customer leads with her team.
“If there’s a lot of manual stuff going on, there’s a tech solution,” she says. “Now I have a partner to create solutions. It’s really saving me from all the mundane tasks.”
Coding has long been limited to the realm of software engineers who studied it in school. But now there are so-called no-code, low-code and pro-code AI tools such as Power Apps, the new App Builder agent in Microsoft 365 Copilot, GitHub Spark, VS Code and GitHub Copilot. That’s broadening who can build apps, helping everyone from non-technical business professionals to experienced developers solve problems, save time and boost creativity.
A productivity edge for engineers
It’s a new mindset that many developers have watched with a welcome but wary eye.
Vibe coding excels with prototyping and getting concepts out of heads and onto screens, helping address what some programmers call “the blank page problem.” AI tools are also invaluable when developers want to experiment or implement improvements that would otherwise require tedious, line-by-line changes — work so time-consuming they might never attempt it.
“The barrier to getting your idea on paper or on a computer or getting it working, getting the first cut, first draft, is down to zero,” says David Fowler, a distinguished engineer at Microsoft who works on cloud native development. “Anyone can do it.”
And coding with AI tools can have a “multiplier effect” for software engineers by taking care of the research and prep work for a project, keeping developers from getting bogged down by details and helping productivity skyrocket, Fowler says.
He’s used GitHub Copilot and VS Code himself to automate monotonous tasks — “stupid things that I repeat daily and are just inconvenient enough that it’s like, I should spend 5 minutes to vibe code an app,” he says — freeing up time and mental bandwidth for more meaningful work. A colleague recently used vibe coding to turn one of Fowler’s apps into a robust website the team now uses daily, all without writing a single line of code, Fowler says.
The power of a quick proof-of-concept
Collaborating with AI changes the focus for developers, he says, by letting them zero in on the bigger picture while the tool works on the coding and architecture necessary to accomplish the goals it’s given.
“It’s outcome-driven development,” Fowler says. “In coding, normally it’s all about ‘how.’ Vibe coding is all about ‘what.’”
With traditional software development, every revision can take a lot of time, and programmers might hesitate to experiment for fear of wasting resources. AI assistance with the basics is having a “profound impact” by helping reduce that effort so developers can try out different ideas and make improvements more quickly, says Amanda Silver, a corporate vice president at Microsoft leading the product team for Apps & Agents as well as a team focused on improving developer productivity.
Vibe coding shows developers what’s possible in a limited form, which encourages them to “do the work to get it all the way to reality,” says Brendan Burns, a corporate vice president on the Microsoft Azure team. “For a lot of people, just getting started is the hardest part. So that’s valuable.”
Spec-driven development: why speed needs structure
At the same time, developers are apprehensive about the new trend for multiple reasons, Fowler says.
Software engineering isn’t “just spitting out code,” he says, but involves gathering requirements, considering the purpose, contemplating how a program might evolve over time, and other “people-centric” activities. And since “software always breaks,” he says, there’s also “a big concern that software engineers are going to have to clean up the messes that are made by vibe coders.”
Maintaining human intelligence is central to coding, says Sarah Bird, who leads responsible AI efforts at Microsoft.
“Vibe coding is exciting for quick proofs-of-concept or experimentation,” Bird says, “but when it comes to building production-grade software, we believe AI-generated code still requires rigorous, specification-driven development.”
Microsoft’s AI tools and coding agents have “protection embedded at every layer,” she says, “combining developer vigilance and continuous testing to catch issues before they reach users.”
That’s necessary because the very speed and ease that make vibe coding so appealing can also lead to shortcuts — like skipping over important security checks or locking in early hacks that become hard to fix later. As projects scale, those shortcuts can create real liabilities, from broken features to vulnerabilities that could hurt a business or its customers.
“If people think we’re going to be able to get to a robust, truly scalable solution just using natural language prompts,” Silver says, “and not understanding the system architecture and the dimensions of complexity of building an at-scale solution — we’re just not there yet.”
Opening the door — without handing over the keys
Guido van Rossum, who created Python, a popular programming language used by millions of developers, says he frequently turns to vibe coding in GitHub Copilot and VS Code in his role as a distinguished engineer at Microsoft, where he’s working on bolstering memory in AI agents. But he sees tools as assistants, not replacements.
“With the help of a coding agent, I feel more productive, but it’s more like having an electric saw instead of a hand saw than like having a robot that can build me a chair or a cabinet or something,” van Rossum says. “I still have to work on the implementation, but the AI help means that trying different things or changing my mind is easier.”
Vibe coding unlocks creativity and speed, “but it really only delivers production value when paired with rigorous review, security and developer judgment,” says GitHub Chief Product Officer Mario Rodriguez. “At GitHub, we’re enabling collaborative, secure and high-quality software creation with Copilot coding agent, which keeps humans at the center while making sure AI augments rather than replaces engineering practices.”
Rodriguez says understanding, testing and verifying code “will remain essential skills — debatably now more than ever — when building with AI in production environments.”
For Ryan Cunningham, who leads Power Platform at Microsoft, AI’s support through new approaches such as vibe coding represents a long-awaited democratization of software creation.
“There’s this inherent tension in this idea of who has access to tech,” Cunningham says. “When is it OK to not be serious about it and to be creative and exploratory? And when is it OK to get serious and build more advanced things? That’s what’s part of why people are talking so much about this idea, because it kind of touches this nerve in a lot of different directions.”
Until now, there was always “a lot of people and expertise and tech in between a person with a need and instructions to computers,” he says. With AI bridging that gap through conversation instead of code, people in diverse roles across an organization can now get involved in projects and help shape tech solutions earlier — and in new ways.
“The lines are really blurring between people who solve problems for the business and people who write software, who used to be very different types of humans,” Cunningham says. “Those groups are really coming together.”
Magic for any age, safety for every stage
Microsoft has been careful to design tools that are safe for those without computer science degrees to experiment with building software, Cunningham says.
“The nice thing in Power Platform is you’re kind of doing it in a sandbox,” he says. “When you push it out to your company, it’s going to happen in the context of our secure boundary. It’s going to run on a platform that is secure and scalable, and all that stuff is handled for you.”
And one “cool thing” about Spark, Fowler says, is that the AI-powered tool “gives you a shareable URL so you can share with friends, and they can log into your app, they can play with it. It runs, maintains and hosts your app on the internet securely.”
Cunningham, Fowler and Silver all vibe code with their children on personal projects, too, that don’t require enterprise-grade scrutiny. They’ve built fun games as well as apps that have helped their kids manage a summertime lemonade stand, track the higher homework load that came with moving up to middle school, or even conquer nightmares.
“Those are the kinds of things you can do in 20 minutes,” Cunningham says. “The ability to tap into that root of exploratory creativity is actually really powerful. It’s magical for anybody of any age.”
For Silver, this new mindset of partnering with AI to build apps is the culmination of a mission she’s been championing for 24 years.
“What I’ve spent my career at Microsoft doing is trying to make it so that more people can create with technology,” she says. “Vibe coding is actually allowing more people to create with technology.”
Illustrations created by Makeshift. Story published on Nov. 13, 2025.
