How to take an AI or vibe-coded prototype to production

Vibe-coding is great for discovery — but turning a prototype into production software still requires real engineering. Here's what we focus on when we make that jump, and how Cut Club came to life.

I'm finally back at my desk after chasing the first week of Tour de France stages from Barcelona to Bordeaux. Tadej Pogačar remains the GOAT and it was another week of 100ºF+ days on this side of the pond. Summer isn't even halfway over and I'm already looking forward to cooler weather!

In other hot news, we've officially launched our first game: Cut Club! After years of playing Music League (thanks, Mike!) we got frustrated with some of its shortcomings and decided we could build a similar, better game.

Cut Club started the way many of our products now start: with a vibe-coded version that let us click around, test the feel of the idea, and decide whether it was worth turning into a production-quality app.

You're probably already familiar with Lovable, Replit, v0, Claude Code, and other vibe-coding platforms. I love these products. They are really great at getting an idea into a usable prototype very quickly. But it's important to understand what they are good at and what their limitations are.

In many cases, your first pass at building something on these platforms makes you feel really optimistic. Lovable is great at building beautiful UIs. It feels like you accomplished 80% of your goal in just a few minutes! But building software is often an iceberg; what you see on the surface is only a fraction of the whole.

Today's vibe-coding tools are genuinely good at generating attractive interfaces, familiar application patterns, CRUD workflows, and convincing happy-path demos. What they do not automatically provide is the engineering foundation required when an application starts handling real users, sensitive data, money, integrations, changing requirements, and complex business rules.

Many of these things are incredibly nuanced and specific to an individual business. Because vibe-coding platforms are powered by LLMs, they are essentially prediction machines, and unless your business is 100% predictable (and if yours is, are you looking for investment?), they will often get things wrong — unfortunately, many of those things are not plainly visible from the start.

When we convert a vibe-coded prototype to a production-ready application, here are some of the areas we focus on:

  • Security and threat modeling
  • User authorization and permission structures
  • Business data modeling and integrity
  • Architecture and maintainability
  • Technical debt and the cost of future requirement changes
  • Testing beyond the happy path
  • Deployment engineering and devops
  • Scalability, performance, analytics, and instrumentation
  • Reliability and failure recovery
  • Compliance, privacy, and data governance

Vibe-coding has dramatically improved product discovery. It lets teams prove an idea before making a large engineering investment. With the tools available today, it is definitely getting easier and faster (and cheaper) to build high-quality, secure, scalable, and resilient software products, but there is still no magic wand.

This is one of the most exciting and fun times to be building software. Our team is now able to build things in weeks instead of months. Writing code has become a commodity, but knowing what code needs to be written and how to design, architect, and organize it is still critical to business.

If you give Cut Club a try, let me know what you think!

Cheers,

Ryan