← All Guides

Your Prototype Shipped. Now What?

Bolt, Lovable, v0, and Replit made shipping your prototype trivial. Then week two arrives: auth fails, data vanishes, deployments break, and nobody answers when it's on fire. The prototype was the easy part. Production is the job.

Key Takeaway: AI code generators are brilliant at validation. They're terrible at production. The moment you need real users, persistent data, secure auth, or someone to call when it breaks, you've outgrown the prototype.

Published: 27 August 2026 • 7 min read • By Paul Gosnell

Need to take your prototype to production?

We rebuild AI-generated prototypes into production systems. Real backends, proper auth, actual databases. 6-day MVPs from £3,000. You own the repo.

Get a quote

The Prototype Trap

You had an idea. You fed it into Bolt. Twenty minutes later, you had a working app. It looked real. It felt real. You showed it to a few people, and they loved it.

So you kept going. Added features. Tweaked the UI. Invited more users. Everything worked.

Then week two hit.

Week Two: When Everything Breaks

Auth stops working

Users can't log in. Sessions expire randomly. Password resets go nowhere. OAuth returns cryptic errors. The prototype used a demo auth provider that was never configured for production.

Data disappears

Your prototype stored everything in local storage or a free-tier database that just hit its limit. Real users expect their data to persist. Your prototype was never built for that.

Deployment is broken

The preview link worked fine. The production deploy is a blank screen. Environment variables are missing. Build steps fail halfway through. The generated deployment config assumed a setup that doesn't exist.

Nobody answers when it breaks

Your users expect support. They expect uptime. They expect someone to fix it when the app crashes at 2am. Bolt doesn't answer the phone. Lovable doesn't debug production issues. v0 doesn't have an on-call engineer.

You can't extend it

AI-generated code is optimised for demos, not maintenance. Components are tightly coupled. Logic is hardcoded. Error handling is missing. Every new feature breaks three others.

Why Prototypes Fail in Production

Prototypes and production systems serve different purposes. A prototype proves an idea works. A production system keeps working when everything goes wrong.

AI code generators are brilliant at the first part. They're terrible at the second.

What prototypes skip:

× Proper error handling and recovery
× Security hardening and input validation
× Database migrations and schema management
× Monitoring, logging, and observability
× Load testing and performance optimisation
× Backup strategies and disaster recovery
× Documentation and handover materials
× Someone to call when it breaks

None of this matters for validation. All of it matters the moment you have real users.

What Production Actually Requires

Taking a prototype to production isn't about fixing bugs. It's about rebuilding with the assumption that everything will go wrong.

Authentication & Authorisation

• Secure session management
• OAuth integration with real providers
• Password reset flows that actually work
• Role-based access control

Data Persistence

• Production database with backups
• Schema migrations and version control
• Data validation and constraints
• Scalable storage for uploads and media

Deployment & Infrastructure

• Repeatable build and deploy process
• Environment variable management
• CDN and asset optimisation
• SSL certificates and domain setup

Monitoring & Support

• Error tracking and alerting
• Performance monitoring
• Uptime monitoring
• Support process for when things break

Rebuild vs Fix: What Makes Sense

The hardest question: do you fix the prototype or start fresh?

Honest answer: if the prototype validated the idea, bin the code and rebuild properly. You'll ship faster and spend less.

Rebuild makes sense when:

✓ The prototype proved the concept
✓ You need features the prototype can't support
✓ Security and data integrity matter
✓ You're planning to scale beyond a few users

Fixing might work when:

→ The prototype is genuinely simple
→ It's an internal tool with low stakes
→ You have a developer who can refactor safely
→ Budget is tight and timeline is flexible

What it costs:

• Simple rebuild: £3,000-5,000 over 6-10 days
• Complex system with backend, auth, payments: £8,000-15,000 over 1-3 weeks
• Fixing prototype code: add 30-50% to rebuild cost and timeline

Frequently Asked Questions

Can I use a Bolt/Lovable prototype in production?

You can try, but most prototypes fail under real-world load. Auth breaks, databases lose data, deployments go blank, and there's no support when things catch fire. Prototypes validate ideas. Production systems need architecture, security, monitoring, and someone who answers when it breaks.

How much does it cost to take a prototype to production?

Depends on complexity. Simple apps rebuilt from scratch: £3,000-5,000 over 6-10 days. More complex systems requiring proper backend architecture, auth, payments, and integrations: £8,000-15,000 over 1-3 weeks. If you're trying to salvage prototype code rather than rebuilding, add 30-50% to those numbers.

Why can't I just fix the prototype myself?

You can, but AI-generated code is rarely structured for extension. What looks like a simple auth fix reveals tightly coupled components, hardcoded assumptions, and missing error handling. You'll spend more time untangling generated code than rebuilding properly.

What do I keep from the prototype?

The validation. You proved the idea works, identified what users actually need, and demonstrated enough value to justify building it properly. Keep the learnings, the design direction, and the user feedback. The code itself is disposable.

How long does a production rebuild take?

For a validated prototype with clear requirements: 6-10 days for simple apps, 1-3 weeks for complex systems. We ship faster than most agencies because we're rebuilding a proven concept, not discovering requirements from scratch.

Ready to Take Your Prototype to Production?