Honest Review: Lovable 3 Months Later

I built six apps in Lovable over the last quarter. Two made it to production. Four didn't. The pattern of which made it and which didn't tells you when Lovable is the right tool.

What Lovable nails

Polished first versions. The first version of any app from Lovable looks good. Better than I could ship in a day of pure coding. The choice of Next.js + Supabase + Tailwind + shadcn is the stack most teams would inherit happily.

Iteration speed for new features. Adding a new screen or new database table is fast. The chat-driven interface stays useful through the first 10-15 iterations.

Database scaffolding. When the project needs Supabase tables, Lovable creates them with sensible columns and relationships. The migrations are clean enough to inherit.

Authentication and routing. The boring infrastructure is handled. Auth flows work. Routes are sensible. You don't waste a week on basics.

Where Lovable struggles

Complex existing apps. After 20-30 iterations, the codebase has accumulated patterns the AI starts inconsistently maintaining. New features sometimes break old features. The chat interface starts losing the plot.

Custom integrations. When you need to wire a specific third-party API that's not in Lovable's common patterns, the chat-driven approach can produce wrong code. You usually have to fix it manually.

Performance optimization. Lovable doesn't think about performance until you tell it to. The default output works but isn't optimized.

Production hardening. Error handling, logging, rate limiting, monitoring — these are afterthoughts that you have to specifically request and review.

The two that made it

Both were small, focused tools with clear scope. One was an internal tool for a coaching client. One was a lead-magnet calculator for a marketing client.

Both stayed in Lovable for the first 2-3 weeks. Both got exported to GitHub and worked on in Cursor + Claude Code from week 4 onward. Lovable was the runway. The other tools were the cruise altitude.

The four that didn't

All four were attempts to use Lovable for the full lifecycle of a real product.

One died because the requirements changed faster than I could iterate in chat. Two died because the AI started losing context and breaking things on every iteration. One died because I wanted features that needed real backend work (queues, webhooks, integrations) that Lovable couldn't reliably ship.

When to use Lovable

Prototype stage. Lovable is excellent here. Get something working fast. Validate with real users.

Internal tools that don't need to scale. A team dashboard. A simple admin panel. An internal calculator. Lovable can handle these end-to-end.

Marketing tools and landing pages. Lead magnets, simple SaaS landing pages, ROI calculators. Lovable ships them well.

Educational projects. Learning what's possible by building something real, fast.

When NOT to use Lovable

Multi-month product builds. The accumulation problem kicks in. Move to a real dev environment.

Anything with complex async work. Background jobs, queues, webhooks that need to be reliable. Use Inngest or Trigger.dev wired into a normal Next.js project.

High-traffic production work. The performance and error-handling gaps become real. Move to a real dev environment.

My actual workflow

I now use Lovable as a sprint-zero tool. Week 1: build the prototype in Lovable. Week 2: get user feedback. Week 3-4: export to GitHub and continue in Cursor + Claude Code.

Lovable is best treated as the fastest path to "something working." Not as the final platform.

The thing nobody mentions

The "export to GitHub" path is improving. A year ago, exporting from Lovable produced code that took 2-3 days to clean up before you could continue elsewhere. Today it's much closer to ready-to-go. In a year I expect it to be seamless.

The path that wins is going to be: build in AI tools, continue in your own dev environment. The tools that nail the handoff win. Lovable is closer than competitors.

What changes in 6 months

Lovable is racing to close the gap on long-iteration consistency. If they get there, the calculus changes — Lovable could be the whole platform for many projects.

Right now, treat it as a runway. Useful runway. Not the destination.