Lovable
In one week, I used Lovable to turn a semi-manual data pipeline into a full micro-SaaS: user signups, credits, pipeline triggers; the works.
I didn't plan on this. I have Lovable Pro through Lenny's Newsletter and thought I'd use it to quickly spin up a React frontend.
Instead, I found a fully integrated platform for building full-stack apps. Thanks to the Supabase integration, the Lovable agent can create and manage databases, implement authentication (Google, GitHub, etc.), run business logic with edge functions, and store files in bucket systems. Frontend and backend get built together, efficiently.
The Lovable agent is primed to work with Supabase, so you don't have to choose an external backend service and spend time integrating it into your LLM via prompts and MCPs.
Beyond Supabase, Lovable has more integrations: Stripe for payments, Resend for email campaigns, Make for automations. If I were a business, I'd be working hard to get integrated.
Want to create an AI app? The agent asks for your API key via a secure handoff, then spins up a Supabase edge function to run inference.
On top of this, you get automated performance and security reviews. The system runs tests, finds issues that need solving, and the "fix it" button sends a prompt to the agent with all the context needed.
When you're ready to publish, they provide hosting. I had my own domain, and when I inserted it, Lovable detected where I bought it and tried to update the DNS records automatically. (It failed, but I love the intention.) Once deployed, they even provide user analytics.
It's all about INTEGRATION. Lovable is solving the "IKEA problem" for app development. When Karpathy wrote about his experience vibe coding an AI app, he noted that "Building a modern app is a bit like assembling IKEA furniture. There are all these services, docs, API keys, configurations, dev/prod deployments, team and security features, rate limits, pricing tiers [...] I spent most of [my time] in the browser, moving between tabs and settings and configuring and gluing a monster."
We're not bottlenecked by model intelligence, but by integration barriers!
Lovable is solving this (Of course, it is only appropriate that the Swedes should be solving the IKEA problem.) But none of it would matter without intelligence and cost-effectiveness. I found the intelligence quite good (most requests were one-shotted), although Sonnet still struggles with frontend and design.
In terms of cost, building the SaaS took 150 messages and ~$18 in Lovable credits. This is so much better than Replit. It's still a bit costly if you're developing intensively, but they don't box you in: you can always get the GitHub repo and work on the code independently.
To be clear: you can't yet pull off a full production app without being technical. There are still planning requirements, complex bugs, and cost/performance/security considerations. But the direction is clear.
(And yes, they just released the Cloud & AI solution yesterday. But this post is long enough)