Skip the coding course today
Monday, 10am. A client pinged the WeChat group for a quote filter page. I stared at a blank doc for ten minutes.
That "I know what I want, I just can't build it" stuck feeling — I've been there. This time I didn't sign up for a three-month course. I described what I wanted in plain words and shipped something clickable and demoable first.
What Vibe Coding actually does
Plainly put, Vibe Coding is briefing AI like you'd brief a new teammate: what's being built, how I'll use it, what comes first. AI turns your description into a web page or small tool. You review the output, call out what's wrong. It's not a one-click delivery button — permissions, privacy, and maintenance still need your eyes.
The sspai editorial crew has been sharing small experiments like this too. My friend Xiaolin (anonymized for privacy) was at her desk in Hangzhou last Thursday at 3pm. She spelled out "filter quotes by budget, flag priority clients, remind me to follow up" line by line — 20 minutes later, she had v1. I also messed this up earlier: I just said "build a client tool" and the AI mashed up pages, login, and reminders together. Once I broke it into three separate actions, things got stable.
If you're replicating today, run the numbers first
Money: Free tiers can start at $0. On a personal plan, I'd cap it at $3—$15/month to start.
Time: About 30—60 minutes for a demoable v1, half a day for simple tweaks.
Technical barrier: You don't need to write code — but you need to articulate clearly who uses this, what they input, what they should see.
First step: Using Lovable as an example: hit "Create a new project" on the page, then click the input box and send your first sentence of requirements.
I'd start with a read-only page: price filter options, a results area, and two example rows. Don't connect real client data yet — wait until you've confirmed it's actually useful. If you don't try right now, that's fine too. Not everyone needs this tool.
Pick by where you are right now
If you're just starting: I'd only build the smallest possible version — that one quote filter page. Goal: get something to the client by tonight.
Got 1—2 clients: I'd book a quick call and watch them click through it. Log where they get stuck, what they misunderstand, what they skip. Then iterate on v2.
Currently scaling: I'd turn recurring requests into templates, and ask someone who knows security to check login permissions, client data, and payment records.
The safest order: prove someone wants it first, then decide if it's worth building long-term. Being able to generate something ≠ people will actually use it. Sustained use is the real answer.