I put Google’s vibe coding platform to the test - A PM’s take on what Firebase Studio gets right (and Wrong)
Lately, I’ve been seriously impressed by Google, especially with the performance of Gemini 2.5. While Google may have arrived late to the GenAI race, it's now punching above its weight: outperforming most foundation models and doing it at a significantly lower price point.
That’s why I was bullish about exploring Firebase Studio. Google’s “vibe coding” platform. Think of it as a cloud-based, AI-powered environment that integrates Project IDX, Genkit, and Gemini AI. It promises the ability to prototype, build, test, and deploy full-stack AI apps directly from your browser. No local dev setup. No terminal windows. Just prompts.
Sounds perfect for a non-developer like me, right?
🧪 My Test: Turning My Recipe Blog into an AI-Powered App
I decided to rebuild ankithagadagrecipes.com into a recipe app one that uses AI to recommend and surface recipes based on previous posts.
To be clear, Firebase's UI offers two different user experiences. One is clearly for a developer use-case, where you can upload an existing app's repo, integrate with your app's framework and/or create a workspace from scratch.
My use-case is from a non-developer perspective.
That said, here’s where Firebase Studio shined:
A conversational, prompt-based UI
Integration with Gemini, to enable recommendations
It auto-generated a Gemini API key (👏 no managing or updating keys!)
It deployed the app to the cloud without any infrastructure configuration
Even my vague prompt:
“Build a recipe app based on the posts and recipes shared here: https://ankithagadagrecipes.com”
yielded a solid blueprint with the right features, style guide, and even a Gemini-based AI recipe suggester.
Blueprint for the app - created from a single prompt
But that’s where the good news ends.
🧱 The Walls I Hit
While Firebase Studio promises a 'vibe coding' experience with its intuitive prompts and AI integrations but the core tension lies in its persistent focus on developer-specific actions. For a truly non-developer experience, the expectation to 'upload an existing repo' can still present a barrier, indicating that the 'vibe' is more about simplifying developer tasks than truly abstracting away the developer environment.
Despite being designed for ease of use for the non-developer use-case, Firebase Studio made some assumptions that limit its usefulness for non-developers:
❌ Image Uploads Were Limiting
The platform only accepts manual image uploads. No support for uploading folders, archives, or linking external image; a blocker when trying to replicate a content-heavy site like mine.
❌ Incomplete Content Extraction
Even with additional prompting and direct links to individual posts, it failed to ingest all ~50+ recipes from my blog. Instead, it scraped just a few... some of which weren’t even mine. When I tried to nudge it with another URL, it asked that I provide each post’s URL manually. That’s not scalable.
❌ No Integration with WordPress
Firebase Studio doesn’t yet support connecting to common CMS platforms like WordPress. A simple integration could’ve allowed me to:
Import posts via WordPress API or XML export.
Automatically extract image/media URLs.
Populate recipe data into structured components.
💡 What Would Make This Better
If Google wants Firebase Studio to truly support non-developers, here are some ideas:
Enable bulk media imports through chat interface (e.g. zip, folder, or XML).
Allow other file types beyond just images (including text, CSV)
Integrate directly with CMS platforms like WordPress, Shopify, and Squarespace.
Offer schema-based ingestion tools, so users can define the type of content they want parsed from a URL (e.g., blog post → recipe card)
🧭 Final Thoughts
Firebase Studio shows a lot of promise, especially when compared to other “vibe coding” tools like Lovable. But it still feels like it was built primarily by developers, for developers. As someone who sits between tech and product, I believe the magic of GenAI is in abstracting complexity and Firebase isn’t quite there yet.
If Google truly wants to democratize app development, it needs to go beyond table-stakes of prompt-driven UIs and build deeper integrations that meet users where they are.
Still, it’s an exciting step. And I’ll be watching closely as the platform evolves.