Lovable Ain’t Shit (If You Already Know How to Build)
Someone asked me a few weeks ago what I thought of Lovable. I started typing an answer, deleted it, started over, deleted that too. Because my honest take requires a disclaimer up front: I am constitutionally suspicious of new build platforms. It’s basically a character flaw. If you’d handed me a shiny app builder in 2002, I’d have immediately asked who owns the export button and what happens when the pricing changes. Grumpy old web guy, twenty-plus years of scar tissue, opinions firmly in place.
With that said: Lovable ain’t shit. At least, not if you already know how to build for the web.
Credit where it’s actually due
Here’s the thing: Lovable is genuinely impressive for what it is. They went from launch in late 2023 to $500M in annualized revenue and a million new projects a week as of June 2026. A $6.6B valuation by late last year. That growth is not a fluke.
The number that actually clarifies everything: 63% of vibe-coding platform users are non-developers. That’s who this is built for. People who have a real idea, zero coding background, and need to get something working. For that audience, it genuinely delivers, and I’m not going to pretend otherwise.
Where it stops making sense
The problem is when experienced builders start treating it like their primary tool. I don’t see the value there, not for anyone who already knows how to design and ship for the web.
The designs it produces are… fine. Generic. To my eye, there’s nothing distinctive coming out of it, nothing that signals a real design decision was made. You can get something on screen fast, but “fast and forgettable” isn’t much of a pitch when the thing represents your business.
Then there’s the credit model. Pro is $25/month for 100 credits, which sounds reasonable until a complex generation burns 2-3 credits and a real production feature runs you 30-60 credits. That’s $40-60 per meaningful feature, with credits that expire after two months. Reddit is not quiet about how this plays out in practice. It’s a treadmill, and someone else controls the speed.
Another walled garden, just shinier
I was skeptical of Webflow years ago, and the concern was the same: the platform you build on top of can change the deal whenever it suits them. And they do.
Unity announced a retroactive per-install fee in September 2023. The industry revolted, the CEO left, the fee was cancelled, and the trust never fully came back. Webflow has restructured pricing twice since late 2024, with freelancers and agencies petitioning each time. Framer users have had their own version of this story. The pattern is not a coincidence.
To Lovable’s credit: the code is real React/TypeScript and it syncs to GitHub on all plans. Genuinely better portability than Webflow ever managed. But the workflow isn’t portable. There’s no import path for an existing codebase. The AI iteration loop, the hosting, the Supabase coupling, the credits: all of it lives inside their walls. The code being exportable is real, and it matters. It still doesn’t stop them from restructuring everything around it.
A few things worth knowing about the track record
I’m going to state some sourced facts here, not to pile on, but because they matter if you’re thinking about building real things on this platform.
In April 2025, Guardio Labs ran a benchmark testing how easily vibe-coding platforms could be prompted into generating phishing pages. Lovable scored 1.8 out of 10, worst of the platforms tested.
A month later, researcher Matt Palmer disclosed CVE-2025-48757: 170 of 1,645 scanned Lovable apps had database tables readable by unauthenticated users, with missing row-level security exposing emails, payment data, and API keys. Per Palmer’s disclosure, the company’s position was that securing app data is the developer’s responsibility.
In April 2026, a bug left chat history and source code from public projects accessible to any logged-in user for about 2.5 months. A bug report filed March 3rd sat marked as a “duplicate” for 48 days. Once a researcher escalated it publicly, the CEO owned it directly and they patched it in two hours. I respect the transparency when they finally moved. The 48-day wait is the part I can’t shake.
What I’d tell a builder who already has the skills
Do the thing yourself. You have the skills. Use them.
Claude Code paired with Google’s DESIGN.md spec, which is a single file that encodes your whole design system as tokens plus rationale that AI agents can consume, produces meaningfully better output than Lovable’s defaults. Stack Anthropic’s open-source frontend-design skill for Claude Code on top of it, which pushes the model toward committing to a real aesthetic direction instead of generic AI output, and you’re building something with an actual point of view. My entire 2026 development workflow runs this way, and I genuinely like how it’s going.
Then deploy wherever makes sense for the project. I lean on Cloudflare for a lot, and most people leave a ton of that stack on the table. It’s still a platform (I’m not going full “own every bit of infrastructure” purity here), but nobody’s going to 4x your bill overnight or sunset the feature your whole workflow depends on. You get the benefit of AI-assisted building without renting your entire process from one vendor.
It’s the same argument I made about managed hosting: the platform isn’t doing what you think it is. Different product, same pattern. The comfort of handing it off, and the real cost of what you give up to do it.
Tell me where I’m wrong
I know I’m the guy who’s been side-eyeing platforms since before some of you were doing this professionally. I have a prior! Maybe I’m missing something real about Lovable that would genuinely change my mind. If you’re a skilled builder getting distinctive value out of it that you couldn’t get building with a coding agent directly, drop it in the comments. I’m actually curious, not just looking for a fight.
