The End of Boilerplate: How AI is Changing the Way We Build Side Projects

Remember the "good old days" of starting a new side project?

You’d spend the first weekend just setting up the environment. You’d configure Webpack (or Vite), set up ESLint and Prettier, wire up a basic CI/CD pipeline, configure the ORM, and write the initial folder structure. By the time you actually wrote a line of feature code, your motivation was already at 20%.

We called it "boilerplate fatigue."

Fast forward to today. I can open ChatGPT, paste a prompt like "Generate a Next.js 14 boilerplate with Tailwind, Prisma, and NextAuth," and within 30 seconds, I have a fully functioning repository.

AI has fundamentally changed the way we start projects. But it’s also changed the skills we need to be successful.

The Shift from "Typing" to "Architecting"
When AI can write your boilerplate, your API routes, and your basic CRUD operations, the bottleneck is no longer typing speed. The bottleneck becomes system design.

If I don't know how to structure my database, AI will just guess (and often guess wrong). If I don't understand the difference between Server Components and Client Components in React, the AI will generate a mess of hydration errors that I won't know how to debug.

AI doesn't replace the need for fundamental knowledge; it amplifies it. It allows you to skip the boring parts, but you still need to know how the engine works to drive the car.

The New Side Project Workflow
Here is how I build side projects now:

Design the Schema First: Before touching any AI tool, I design the database schema and the core user flow. If the foundation is wrong, the generated code will be a house of cards.

Generate the Scaffolding: I use AI to generate the initial scaffolding, UI components, and basic API routes. (I keep a vault of AI prompts specifically for developers to get the best output).

Review and Refactor: AI code is often "smelly." It works, but it might not be efficient or scalable. I spend my time reviewing the generated code rather than writing it from scratch.

Ship Fast: Because the boilerplate is done in hours instead of weeks, I can actually get to the deployment phase in a single weekend.

The Danger of Over-reliance
The biggest trap developers fall into right now is accepting AI-generated code blindly.

I recently asked an AI to write a complex SQL query. It wrote a beautiful, perfectly formatted query — that used a deprecated syntax and would have locked up the database under heavy load. Because I understood the underlying SQL, I caught it. A junior developer might have pushed it to production.

Use AI as a junior developer. It’s great for writing the repetitive stuff, but you are the senior. You are the one who signs off on the pull request.

What's Next?
We are entering an era where the barrier to entry for building software is lower than ever. But the barrier to building good software is still just as high.

If you are starting a new project this weekend, don't spend hours tweaking your ESLint config. Prompt the AI, generate the base, and focus on the features that actually matter.

If you want to launch your next side project but hate setting up the boring infrastructure, I've packaged my exact step-by-step checklist, Notion workspace, and deployment guide into a Dev Launch Kit. It goes from zero to deployed in a single evening.

And if you want to get the most out of your AI coding assistant, check out the AI Prompt Vault — it has over 500 prompts specifically tailored for developers.

What's your take? Do you use AI for boilerplate, or do you prefer to write it yourself? Let me know in the comments!