Plan before you prompt

In the last lesson, you used Plan mode without much explanation. Lovable thought before it built, you approved the plan, and the card came out right on the first try. Let's talk about why.

You type: "Build me a feature request board with voting, user accounts, categories, admin responses, search, filtering, and a nice design." Lovable generates something. But one prompt can't carry that much intent. The AI has to guess at every detail you left out, such as layout, priority, how features connect, or what "nice" means. You get something 60% right in eight places and 100% right in none. You send five follow-up prompts, but each one shifts the foundation because the foundation was never defined. By prompt eight you've spent more time correcting than building.

The fix isn't a clever prompt, but rather a plan. And you already used one. In the last lesson, you switched to Plan mode, described a card, reviewed the plan, and approved it before Lovable wrote a single line of code. That worked great for a single card. Now we're going to use the same mechanic for something bigger, and you'll see why it really matters.

You don't create a master plan for the whole project. You break the work into chunks and plan each chunk right before you build it. Each plan does its job the moment the code ships. They're saved in your project under the .lovable/ folder if you ever want to look back.

We know the first thing Upvote needs is a skeleton, which is the basic page layout that everything else will live inside. Switch to Plan mode and type this:

Lovable might come back with a plan right away, or it might ask you questions first. Things like: how many categories do you need? Should the sidebar be collapsible? Do you want sorting controls above the list? Answer them, and the plan that comes back is sharper because of it.

When you're happy with the plan, click "Review." You'll see it as a document you can read through and edit directly. If something looks off, change it by typing in the document. When it looks right, approve it. Lovable switches to Build mode and starts building. Let it run, and we'll look at what it built in the next lesson.

After the skeleton is built, you'll use the same loop for the next piece. Need a submission form? Flip to Plan mode and describe what it should do. Lovable thinks it through, maybe asks a few questions. You review, approve, and it builds. Then move to the next chunk, and so on.

This means Plan mode isn't a one-time ceremony at the start of a project. It's something you reach for whenever the next step feels big or ambiguous. Three lessons from now, if you want to rethink how voting works, flip to Plan mode: "Here's how voting works now. I want to add downvotes and a score system. Help me think through the implications." Same loop, but smaller scope.

Here's the insight worth internalizing: planning doesn't slow you down. It actually speeds you up. Without it, you waste ten prompts wandering. With it, every prompt has a clear purpose.

And think about this before we move on. If you had to build Upvote one piece at a time, what would you build first? You might be tempted to start with voting or the submission form which are the interesting parts. But you build the skeleton first. Think of it like the empty rooms of a house. Get the walls and hallways right before you bring in furniture. If you rearrange the floor plan after everything's decorated, you're moving every piece twice. That's exactly what we planned first, and it's what is building now.