Vibe coders build tiny apps instead of chaos

vibe coding – A growing group of non-traditional programmers is using AI to create working, hyper-specific apps—everything from wedding seating plans to grocery deal tracking—after coding-focused models made the process more reliable. For them, the appeal isn’t replacing wo

Every time Juliana sees someone open ChatGPT on the subway or in a grocery store, a familiar dread creeps in: the feeling that basic questions could be answered with a bit of thinking—or with a search that people actually evaluate.

But when she was assigned to look into the rise of “vibe coding. ” a practice where non-coders use AI tools to build apps. she ran into something different from the brain-rot story she expected. She found hobbyists—people of all ages and professions—treating app-building like a playful craft. They weren’t chasing a new economic order. They were trying to make everyday life easier, one small working program at a time.

In college, Shayan Mirzazadeh failed computer science twice. A decade later, the 31-year-old account manager uses vibe coding to build solutions to pain points at work and home. His projects include an app to help his fiancée track her Pilates flows.

His co-worker and fellow “side quester. ” Jayne Ingram-Roberts. built a fantasy-league-style app for the TV show “Big Brother.” Their biggest and latest project is called “Seatbee. ” a website for crafting wedding seating plans—an idea Ingram-Roberts says grew out of her own trouble planning her wedding two years ago.

On Seatbee, users input rules for who should and shouldn’t sit by each other. Ingram-Roberts describes the stakes in personal terms: it’s “super important that all my work friends sit together. ” and it’s “important that my sister and this drunk uncle are on opposite sides of the room.” Once the rules are in. users click generate and a seating plan comes out. The duo says the website already has over 200 users.

That level of specificity—far from building “enterprise” software—points to the core reality of vibe coding. These AI hobbyists aren’t trying to blow up the world with general-purpose platforms. They’re building apps for singular moments and narrow needs. and. like many hobbies. they’re happily losing money along the way because the fun is part of the point.

The shift didn’t happen instantly. Longtime coder and writer Paul Ford dates the vibe coding “vibe shift” to November 2025. Before then, Ford says AI might produce a webpage, but the quality was inconsistent. It couldn’t necessarily debug itself as it went. AI coders still needed hand-holding.

The tipping point. he says. came when models started writing code. running it. identifying bugs. and then debugging on their own—seeing the task through. That progress was “turbocharged. ” Ford says. by the launch of coding-focused models from major AI companies. including Anthropic’s Claude Opus 4.5. Google’s Gemini 3. and OpenAI’s GPT-5.1.

For Ford, the key difference is whether the output works.

“With coding, it was like, no, actually this works, and the computer doesn’t care whether it’s slop or not, it runs it,” Ford says. He adds that vibe coding rests on a baseline that prose or art doesn’t have. “That really is this baseline that doesn’t exist when it makes prose or art,” he says.

And unlike run-of-the-mill chatbot usage—which can worsen long-term critical thinking or deskill workers who use it to power through the day—vibe coders aren’t treating AI as a one-to-one brain substitute. They use it to do something they couldn’t do before, while learning in the process.

One study described vibe coding as similar to making pottery: the manipulation of the material—whether it’s clay or code—is “inseparable from the potter’s thinking and the development of skill and intention.” Rather than asking a chatbot for a fully formed product. iterating on code with an AI model is compared to guiding a robot’s hand as it shapes clay: figuring out when to stop. and when to tell it to adjust.

Jonathan Butler, a 56-year-old entrepreneur and vibe coder, used to rely on other people to help build websites. He compares learning to another craft. “It’s like being in your wood shop making something,” Butler says. His newest project is vibe coding the construction management process for his new home.

The appeal also runs up against how many enterprise apps are built. Enterprise software often aims for scale, adds features that can alienate users, and sometimes hides favorite tools behind paywalls. Vibe coding. by contrast. starts with one small problem and produces an appropriately scaled solution instead of building a blunt instrument and then winnowing it down.

That approach is part of why the practice is showing up beyond traditional developer circles.

Software engineers Maya Miller. 28. and Chloe Garden. 29. run the SiSTEM Collective. a New York-based group for Black and Latina women working in tech. Part of the group focuses on community-building workshops for people who are less tech-savvy. A recent workshop centered on vibe coding: women came in with an app idea and left with a working prototype. Miller and Garden describe turnout of around 30 people. including five or so complete amateurs and about a quarter of attendees who had never really drilled into the nitty-gritty.

For newcomers—and for those with a bit more experience—vibe coding lets them address specific day-to-day problems.

Miller gives a concrete example: “Two people were working on hair wash day routines — so just tracking the products. if it’s actually helping their hair goals. if they want to grow length or increase the elasticity of their hair.” She calls it a case of “bespoke software that wasn’t really accessible.”.

The metaphor they return to is practical. It’s “the virtual version of finally fixing that leaky sink,” Miller says in the story’s framing. Or setting up something like a routine for who feeds the cat so it doesn’t end up “swindling its way into a double-dinner.”

Scott Klipper built an app to help with his kids’ school pickup. Joe Poynton. a firefighter. used vibe coding to save a few moments at the grocery store by using a list that sorts items by location. The promise, in this telling, is not that software will automate life completely. It’s that it can step in to solve little problems: telling you. based on your workouts and brunch plans. whether it’s a good day to wash your hair.

None of it guarantees bigger change. Vibe coding “won’t magically resolve the structural and societal questions that come with the AI boom,” the story notes. But it does offer real-world examples of how AI can work for an average person. People can solve their own problems—or at least create micro-apps that won’t suddenly be “enshittified” by new paywalls and bloated features.

What also surprised Juliana, after talking to vibe coders and AI hobbyists, was how much they love the game. They aren’t trying to dominate app stores or become billionaires. They aren’t building endless customer-service bot loops designed to annoy and extract money. “They’re in it for, literally, the vibes,” the reporting says.

The feeling echoes an older internet: a more decentralized, bespoke version of the web, where people made programs devoted to sending a friend the word “Yo,” or “drinking” a virtual beer.

Kyle Jensen. a developer and director of entrepreneurial programs at Yale’s School of Management. recently vibe coded an SAT prep app for his kid. He also made research apps for his wife and search apps for his colleagues. In his circles. Jensen says interest is exploding—though he admits the group is made up of people studying management or otherwise attuned to software.

Even so, he points to something larger: services that could help more laypeople embrace vibe coding. “That probably points to a future where normies, regular people, are deploying apps pretty regularly,” Jensen says. In his view. it’s about “threading the AI needle.” Paul Ford sees vibe coding as one way to give the masses more control over AI—especially for people who don’t want to talk to robots all day but want to build something or get work done.

To close the story, Juliana tries it herself.

She decides to solve a specific New York City problem: she has two grocery stores within walking distance, but one is up a steep hill. She wants to compare weekly deals to decide when it’s worth the uphill walk.

Her result is mixed. She learns that it’s “still very hard to scrape weekly circular PDFs”—grocery stores are “accidentally AI-resistant.” She says she learned a lot about how scraping works and how systems code can decode information from “archaic supermarket sites.” She also calls out how effective prompting becomes part of the work: she began by treating her Codex like a fellow reporter or editor brainstorming an app. but it “wasn’t capable of understanding that level of nuance.”.

Even when the initial plan fails—PDF circulars defeat her—she pivots. “Ultimately, my app wasn’t as successful as I hoped,” she writes. She ends up building an app that weighs whether a deal is worth the extra walk to get an item at the higher-altitude supermarket. She calls it silly and expects people to tease her about it. Still. she says she’ll use it—her discount mozzarella. “and I will be laughing all the way down the hill.”.

The bottom line from the reporting is straightforward: for these builders, the point isn’t to replace thinking with prompts. It’s to use AI as a tool to make small, working solutions—and to keep the process human enough that it feels like play, not production.

vibe coding AI coding micro-apps non-coders wedding seating plan grocery deals Anthropic Claude Opus 4.5 Google Gemini 3 OpenAI GPT-5.1 debugging SiSTEM Collective workplace productivity