Vibe coding from a blank page feels like the purest form of the craft. You describe what you want, the model writes it, you iterate, you ship. Nothing gets in the way. It is also, for most businesses, the wrong way to start.
Here is the simple rule I use.
Start from Scratch When the Idea is Genuinely New
If the thing you are building does not already exist, then yes, start from scratch. Fire up Claude Code, describe your dream, and build until it works.
Start with the blank page if you are creating:
- A novel marketplace for something weirdly specific.
- A collaborative tool with a core loop that has never been tried.
- A developer tool that does not fit an existing pattern.
In these cases, prior art is a constraint, not a benefit. Every existing piece of software was built on assumptions that may not apply to you. Inheriting code means inheriting assumptions that can actively mislead your development.
Start from an Open-Source Base When You are Automating the Known
Almost every business idea is not genuinely new. It is a variation on a known pattern: manufacturing, retail, services, wholesale, or distribution. The process is old. The domain model is old. What is new is your twist on it, perhaps a niche audience, a specific integration, or a new pricing model.
In this case, starting from scratch is a tax. You will spend months re-deriving the chart of accounts, the purchase-to-pay flow, the sales pipeline, the inventory model, and the permission system. This is plumbing that thousands of consultants have already argued about, refined, and converged on.
Meanwhile, a business that starts on Odoo gets Best Practices convergence on day one. They start with:
- An inventory model that handles lots, serials, landed costs, and negative stock.
- An access-rights framework that works across fifty different modules.
- A chart of accounts that actually allows you to close at month-end.
You get to skip the plumbing and spend your energy entirely on the part that is actually yours.
The Winning Strategy: Combine the Two
The best teams I see are not choosing. They are combining. This is the new stack that wins.
- Step 1: Start on an Open-Source Base. Whether it is Odoo, WordPress, or Supabase, choose the base that fits. You inherit best practices immediately.
- Step 2: Vibe Code Your Uniqueness on Top. Use AI coding tools like CodeCobra to build your specific workflow, your custom integration, or your customer-facing surface. Build it as modules or extensions on the shoulders of that base.
- Step 3: Keep Updating the Base. Because the community keeps improving the foundation, you get a better platform for free with every release.
This posture is fundamentally different from writing everything from scratch. It is also fundamentally different from the old enterprise playbook of paying a consultancy seven figures to configure a closed-source ERP. Neither of those strategies win anymore.
The Question to Ask Before the First Line of Code
Before you open Cursor or Claude Code or your favorite IDE, ask one thing.
Is the problem I am solving genuinely unprecedented, or am I automating a known process with a new wrinkle?
- If unprecedented: blank page. Go.
- If known process: find the open-source base that already encodes the convergence. Adopt it, then vibe code your wrinkle on top.
The worst possible path is to start from scratch on a problem that has been solved a thousand times. The AI will happily help you rebuild it. You will ship a mediocre copy of something that already exists in better form, for free, maintained by thousands of people.
Do not rebuild what ten million businesses have already agreed on. Inherit it, then go build the part only you can build.
