totally agree with the idea of a SaaS boilerplate to ship faster; i've been there trying to bolt auth and billing onto a prototype and it kills momentum. first, map the single core user journey you'll validate on day one - signup, upgrade to paid, the primary action, and the confirmation - then only build the pages and API routes needed for that path. keep integrations stubbed behind feature flags so you can demo and run end-to-end smoke tests before wiring everything, and reuse a basic pricing plus invoice flow so you don't invent that wheel. fwiw i used a starter kit that had auth and payments preconfigured (Vibe Coding Starter Kit) and it saved me about a week of wiring Stripe and email, letting me focus on the product UX.
Great discussion. I've been building in this space and think the truth is somewhere in the middle.
The real value isn't in "vibe coding" everything from scratch - it's in having the right architecture and patterns that AI can work within. That's why I built VCSK (Vibe Coding Starter Kit) - it gives you a solid NextJS foundation with proven patterns, then you can vibe code features on top of that foundation.
The difference is architectural discipline. When you start with proper auth, DB setup, deployment pipeline, and component structure, AI becomes incredibly productive for building features. When you ask AI to architect everything from scratch, you get the dead-end architectures you mentioned.
I still write the critical business logic by hand, still review every PR, and still understand the codebase deeply. But for UI components, API endpoints that follow established patterns, and data transformations? AI is genuinely faster than manual coding now.
The key insight: "vibe coding" isn't a replacement for engineering judgment - it's a force multiplier when you have the foundation right.
i did a similar audit and the worst offender was duplicated config and generated files across the Next.js frontend, NestJS backend, mobile wrapper and CI. make a tiny MVP baseline you can actually ship from: one Next app, one thin backend service with a single auth route, and move CI/mobile scaffolding into a separate repo or a generator so they don't come along to every new project. consolidate shared types and infra code into a single package, kill unneeded dev-tooling files, and change heavy codegen to on-demand scripts that run when you add a feature. build a tiny CLI that scaffolds a feature folder instead of including every possible file in the base. that cut my new-project file count from hundreds to a few dozen. fwiw i used a starter kit that had auth and payments preconfigured, Vibe Coding Starter Kit, to get a production-ready baseline without rebuilding all that infra.
This vibe coding concept is pretty interesting. I've been finding that having a solid starting point really helps with keeping everything deterministic. I've been using the Vibe Coding Starter Kit for launching projects quickly and it's been a game changer.
As a developer of more than 15 years, it was a bit of a struggle when AI came in and wanted to change the traditional ways of building products. The results weren't very convincing initially.
But as things progressed, the code quality quickly improved. It allowed me to try out a "pure vibe coding" approach for a few small projects. There were some very frustrating moments, to say the least, but I got them working and learned a lot along the way.
That's when the idea of building a Vibe Coding Starter Kit appeared.
I built a native desktop app that helps you throughout the whole process:
The Generation: An interactive wizard lets you set project details, pick colors, fonts, and integrations (Supabase, Stripe, etc.) to get a project ready to go.
The Setup: It guides you through installing dependencies, pushing to GitHub, and - crucially - "teaching" Cursor about the project, coding guidelines, and security.
The Build: Scaffolding is just the start. The project comes with specific Cursor commands to help you actually build the product, even if you are not a technical person.
The Planning: I included a Project Planner command that interviews you about the software you want to build, integrates with Figma/Google Stitch, and creates a step-by-step plan for the AI.
My main goal was to remove the repetitiveness when creating new projects and allow non-coders to build their own businesses.
The Demo: I used the kit to build the landing page for the project itself. I recorded the process (edited down to ~33 minutes) to show how it works from start to finish.