I have never written a line of code in my life. Not at school, not at university, not in fifteen years of running a laptop-based business. So when I kept seeing people online build entire working apps by typing plain English into a tool called Cursor, I assumed it was influencer exaggeration. I gave myself one week, a real problem I actually wanted solved, and zero prior experience, to find out. This is the honest account of what happened — including the two hours I nearly gave up.

"Vibe coding" is the term that's stuck for this — you describe what you want in ordinary sentences and an AI tool writes, runs and fixes the actual code, rather than you typing syntax you don't understand. Cursor is one of the tools built specifically for it: an editor that looks like a normal code window with a chat panel down the side, except the thing in the chat panel can read every file in your project, make the changes itself, run the result, and see the real error message when something breaks. I picked it because it kept coming up as the one actual builders use daily, not just the flashy demo tool.

Why I even tried this

I wasn't chasing a hobby project for the sake of it. After testing whether 10,000 steps a day was actually a meaningful target, I wanted something to track my own daily count against my own number and show me a simple streak — not another app with an account, a subscription and a notification asking me to rate it. That's a small, boring, entirely personal problem, which turned out to be exactly the right size for a first attempt.

What actually happened when I opened it for the first time

I installed Cursor, opened an empty folder, and typed something close to: "build me a single web page that lets me log my steps for the day, shows a streak counter for consecutive days I hit 8,000, and remembers everything even if I close the tab." I hit enter expecting a wall of text I couldn't read. Instead, files started appearing in the sidebar on their own, a preview window opened showing an actual working page, and within about ninety seconds I had a box to type a number into and a counter underneath it. I hadn't written a single character.

The thing that actually surprised me

It wasn't the code-writing itself that changed my mind about this — it was the back-and-forth. I'd say "the streak resets even when I hit the target two days running" and it would look at what it had just built, work out where the logic was wrong, and quietly fix it, the same way I'd describe a bug to a person sitting next to me rather than filing a formal report. That loop — describe the problem in plain words, watch it get diagnosed and fixed, check the result — is the entire skill. There's no syntax to memorise if the thing next to you already knows it.

"I kept waiting to hit the wall where you actually need to know what you're doing. Four days in, I hadn't found it — I'd just found smaller walls, and each one moved the same way the first one had."

What I actually built

By the end of the week I had a single page, saved locally, that logs a step count with one tap, keeps a running streak against my 8,000 target, remembers everything between visits, and shows a small bar for the last two weeks so I can see the pattern rather than just today's number. Nothing about it is clever software — it's genuinely the kind of thing that would take a developer half an hour. The point is that I built something I now actually use every morning, for a problem no app store listing solved, without paying anyone or learning a programming language first.

Where it properly went wrong

I want to be honest about the parts that weren't smooth, because the "I built an app in a weekend" posts always skip this bit. On day three I asked it to add a weekly summary, and it rewrote the part of the page that saved my data, quietly breaking the streak count in the process — my numbers reset to zero and I didn't notice until the next morning. Fixing it took one more plain-English message and about three minutes, but it's the moment that would have finished off anyone expecting this to be flawless. It isn't. It's more like working with a fast, capable assistant who occasionally breaks something while fixing something else, and needs telling clearly when it has.

What "vibe coding" actually means, once the buzzword wears off

Strip the name away and it's just this: the barrier to building a small, working, personal piece of software has dropped from "learn a programming language" to "describe a problem clearly and notice when the answer is wrong." That's much closer to the plain-English prompting I already use for building a custom GPT than to anything I'd call software engineering. It sits alongside the rest of the practical AI tools I actually use week to week — it just happens to produce a working app instead of an answer.

The honest verdict: I didn't come out of this a developer, and I don't think a week of vibe coding makes anyone one. What I did come out with is a small tool I use every single morning, built entirely by describing what I wanted and correcting it when it was wrong. If you've ever had a specific, boring, personal problem that no app quite solves, that's the project to try first — not a demo, a real annoyance you already have.

Is it worth trying if you can't code

Yes, but pick a real problem, not a hypothetical one. The projects that go well are the ones where you already know exactly what "working" looks like, because you're the only user and you know the moment something's wrong. Start on the free tier — it was more than enough to build and finish what I made, and it's the same instinct behind every other AI habit I've kept: treat it like one of the automations that quietly save real time rather than a novelty you open once. I later put Cursor head-to-head against GitHub Copilot on a slightly bigger build, which is worth reading before you decide which one to start with. If you want the next honest first-week account before it goes live, join the Brite newsletter.

Questions people actually ask

What is vibe coding?

Vibe coding is the informal name for describing what you want in plain English to an AI coding tool and letting it write, run and fix the actual code, rather than typing it by hand yourself. You steer by describing the result and the problems you hit, not by writing syntax. It's less like programming and more like briefing someone who already knows how.

Do I need to know how to code to use Cursor?

No. I had never written a line of code before this week and built a working, genuinely useful tool by the fourth day. You'll get there faster if you can describe clearly what you want and read well enough to spot when something looks obviously wrong, but neither of those is programming knowledge.

Is Cursor free to use?

Yes, there's a free tier with a limited number of AI requests each month, which is enough to try it properly and build something small like I did. The paid tier, around twenty dollars a month, raises that limit and unlocks the stronger underlying models for bigger or more complex projects.

Is code written by AI actually safe to use?

For a small personal tool that only touches your own data on your own machine, yes, and it's the same code you'd get asking a competent junior developer. I would not ship something written this way if it handled other people's data, passwords or payments without a real developer reviewing it first — that's a different level of risk to a private step tracker.

What's the difference between Cursor and just asking ChatGPT to write code?

ChatGPT hands you code in a chat window that you copy, paste and run yourself, hoping you didn't miss a step. Cursor sits inside the actual project, can read every file, runs the code itself, sees the real error message when something breaks, and fixes it in the same loop. It's the difference between a recipe and someone cooking next to you.