In 2026, “vibe coding” has spread from tech circles to everyday builders. The guide walks readers through choosing a platform, writing prompts, fixing mistakes with modifying prompts, and managing limited daily credits—using a subscription tracker built in les
When the reels start playing, the surprise isn’t just that people are building apps. It’s how fast it looks.
In 2026. instead of cat videos. a growing stream of users is posting “I built an app in four hours with Claude Code!” posts that make the whole process seem effortless. The phrase “vibe coding. ” coined by OpenAI founding member Andrej Karpathy last year. describes using plain English to get AI to write code and build apps. The trend has moved far beyond engineers—so far. even people building to support aging parents’ healthcare needs are jumping in.
Tech executives have helped normalize the hype, too. Companies like Anthropic and Google are drawing crowds with claims that 90% of their company’s code is written by AI. Meanwhile. Business Insider tested tools including Claude Code. Manus. Lovable. Base44. and Cursor—building personal fitness apps. newsroom assistants. and an in-house version of Pinterest.
But the first time the writer tried vibe coding, it didn’t land the way they hoped. Prompting AI in plain English turns out to be learnable—but it also takes practice. So here’s the step-by-step guide, built from that experience.
The first decision is the platform. Several products offer users a number of free daily credits to play with.
For non-techies trying to build simple tools rather than complex software products, Lovable, Base44, and Replit are presented as the best bets. If you’re more tech-savvy, the guide points to Claude Code, Emergent, and Cursor as options that might work better.
The writer chose Lovable for the tutorial. One goal: stay mindful of subscription spending. The plan was to build a subscription tracker—explicitly described as a simple, popular use case.
Next comes the hard part that everyone skips in the viral posts: prompting. “AI is only as good as the prompt it receives,” the guide warns.
The easiest way to get started. the writer says. is to ask an AI tool like ChatGPT or Claude to generate a custom prompt for your app. For the subscription tracker example. the writer asked ChatGPT: “Give me a prompt to build a subscription tracker — I want to track how much I’m spending on streaming service subscriptions. news outlets. music apps. workout apps. etc. I want to be able to see at a glance when each subscription starts and ends.”.
Then the user paste-and-build loop starts. The prompt goes into the search bar, with the setting kept on “build,” followed by pressing “enter.” The guide notes to “hang tight,” because this can take a couple of minutes.
If the first draft looks wrong, that doesn’t mean the process is broken. It means you’re not done prompting yet.
The writer’s first version of the vibe-coded subscription tracker is described bluntly: an all-white background and pie charts left them “deeply unimpressed.” The interface looked boring—like something they would stop using within a week. Even the decision to change the website wasn’t optional for them; they wanted it to meet their needs.
So the fix was a modifying prompt, framed like telling an intern: “Change the colors. Make it more fun.” The guide provides space for a modifier prompt and a regenerated version of the website (with the instruction that you can use ChatGPT, Gemini, or other tools to nail the prompt).
When the app regenerated, the visuals changed dramatically. Instead of clean white, it had tacky neon buttons against a dark background and cliché achievement badges like “Budget Guardian” and “Smart Saver.” That result, the writer says, felt “more like it.”
Once the website looked right, the writer directed readers to press the “publish” button to enjoy the fruits of their—and the AI’s—labor.
For the tutorial’s outcome, the timing matters: it took less than 30 minutes to vibe code a subscription tracker with Lovable.
Then comes the money question that sits underneath every free-credit workflow: how do you maximize free credits without burning through them?
The guide sets the baseline expectation: each prompt takes up credits. After finishing the tracker, the writer had only 0.5 of their five daily free credits left.
To ration those credits, readers are told to toggle between “plan” and “build” modes. In “plan” mode, the tool uses only one credit per prompt. The purpose is to plan features and understand changes before modifying the app.
If daily credits run out, the guide offers the two choices: wait for credits to renew after 24 hours, or upgrade.
Lovable’s cheapest plan costs $25 per month and includes 100 extra monthly credits. Replit and Base44 are also mentioned as having plans from around $20.
There’s also a workaround for people who want to stay on the cheaper side: refine the initial prompt with everything you now know, then feed it to another AI platform like Base44.
The result, as the guide frames it, is vibe coding 101—use plain English, iterate with modifying prompts, and treat credits like a resource you manage, not something that just disappears.
vibe coding AI coding tools Claude Code Lovable Base44 Replit Cursor Emergent Manus Andrej Karpathy Anthropic Google subscription tracker free daily credits