17 Comments
User's avatar
Pawel Jozefiak's avatar

This was a really clean breakdown, especially the “what’s actually special vs what’s just hype” framing.

The part that resonated most: free + generous tokens is a feature, but it comes with a real “you’re the product” trade-off. People talk about AI Studio like it is some pure win over Lovable or v0, but the pricing model changes what kinds of projects you can safely put inside it (personal stuff, client work, anything remotely sensitive). That nuance gets lost in the “Gemini is insane” posts.

Also loved the Stitch → AI Studio flow. That is the most “Google” thing ever: genuinely powerful pieces, but you feel the seams between teams/products. The export step is cool, but it also highlights how messy the “end-to-end” story still is.

And the “Annotate app” bit is interesting because it is kind of the opposite of what most vibe-coding tools optimize for. It feels like Google is betting on batching feedback instead of surgical edits. Not sure which will win long-term, but it is at least a different philosophy (and not just a clone).

I went through a similar detour recently (AI Studio + Gemini 3 Pro + Antigravity vs Cursor), and had the same “surprisingly capable, but the seams matter” reaction: https://thoughts.jock.pl/p/cursor-vs-google-ai-studio-antigravity-ide-comparison-2025

Ibrahim Hanif Shuaibu's avatar

I really love how you break this down, even though am still trying to figure out many things. I am learning web development since 2023, I want to pursue that as a career. Do you think using the AI tools is recommended for me?

Xinran Ma's avatar

Thank you! Yeah, I would definitely recommend you to give them a try—you’ll learn things along the way.

Jay's avatar

It's an interesting offering, but not sure if and how it would integrate with Supabase.

Enjoying Google's generous token count is very appealing. But can't you also access that same level of computing from Dyak, which also lets you access many other models?

Xinran Ma's avatar

For Supabase, I have not checked, but assuming with Firebase is a smoother integration since they are all google products.

For Dyad, I've not tried it either—intrigued by it though. But I suspect that even if it can "access" the models, it doesn't mean that it can use those models as effectively as Google's own tools. Just like Cursor's capabilities with the Claude model is different from how Claude Code's uses the Claude model.

Jay's avatar
Dec 8Edited

I suspect you are right. But accessing different models also brings its own advantages, like easy access to alternatives when you are stuck. And also, it ensures you will not run out of tokens, which is currently a huge problem for most vibe coders.

I'm in a moment of transition myself and would love to read your thoughts about Dyad.

Xinran Ma's avatar

True. I need to check it out!

Rohit Kelapure's avatar

I subscribed after watching your breakdown of desing on Peter's youtube channel.

Xinran keep creating buddy. Nice work!

Xinran Ma's avatar

Thanks a lot, Rohit! Appreciate your kind note and support.

Suhrab Khan's avatar

This is a thorough and practical breakdown. Your exploration of Google AI Studio’s workflows, integrations, and multimodal capabilities makes it clear how designers can leverage it effectively.

Xinran Ma's avatar

Thanks, Suhrab!

Justin's avatar

Nice excalidraw. Indeed it's so confusing even to us who are swimming in it. I'll modify mine to include Gemini mobile vs Gemini web and throw in Mixboard

Xinran Ma's avatar

I have not used Mixboard at all! Thanks for mentioning it. And yes, it’s excalidraw :)

Justin's avatar

Thanks for the walk through. Aligns with my experiences.

Xinran Ma's avatar

Thanks for the note, Justin. Glad I’m not alone feeling this. We’ll see how it evolves—likely more experiments by Google happening at the same time.

Magick Mica's avatar

Post saved! I’m going to read this shortly!!!

Thank you❤🙌

Xinran Ma's avatar

Thank you!