For beginner context start [[Trying To Vibe Code |here]] When I wrote that last post I only had experience with ChatGPT and had just started using Cursor. This time I'm going to talk about using Firebase Studio. ## In the Beginning I'm just trying to understand what's going on in the AI space at this time. My experience with just cursor was kinda exciting yet felt stupid. The only way to get something to go from stupid to normal is to keep trying. This time I decided to try out Google products. I was already kinda using Gemini for ansible things but wondered if Firebase Studio could build my whole app and give me the code? Well it's a yes and no. ## What I liked Firebase Studio was pretty simple. I tried a new app idea (S/O to Keenan). I put in the description that he gave me and added some examples and within about 2hrs I had a functional app. Everything felt smooth and feeling like what I expected so I started to look at the code. That's when I realized I was about to lean on AI a lot. I had never seen nextjs code written and I'm not the best web developer. I was determined to stay positive and decided it was time to add in 3rd party auth and a real backend. It was at that point I started getting frustrated. ## Details of Dislike Let's just get this out the way. My immediate dislike is that Firebase Studio is from Google. I still got beef with Google and I'm going to stand on that until they correct the mess. ANYWAY my real biggest gripe initially was that I couldn't figure out how to download my code. I didn't want things stored on Google Cloud and such. After doing some Google-ing I was able to figure out that the only real way to do things was to connect to Github. I wasn't really into that idea, but I needed to get my code. I wasn't going to host through firebase so I needed to get it locally. After fighting that battle I ended up having a problem with Supabase authentication that I couldn't figure out. I'm no web developer but I know enough about web codes to tell when I'm stuck in a loop. I tried asking pointed questions to get Gemini to help me through things but all I got was "I can't help you". In the end I ended up going back to Cursor because every time I blinked Gemini was telling me that it couldn't help. I was getting way too frustrated stuck in a misunderstanding of redirect loops so I just moved on. I never figured out if Gemini was timing out or what I asked just wasn't clear. I wish there was valid feedback I could move forward with but we are done now. ## So how did this end? Well I got a little staging thing going and deployed to Vercel. I should be releasing this to test to a few people to see what they think about it. I'm not really sure how people are building whole things using AI because honestly computers are dumb. Cursor was still giving me trouble with the auth when I switched over to the point that I just started reading the docs myself. I'd also like if there is no set technology set in the initial prompt if the "system" would explain the why for picking that tech stack. The why could just be like "technology - \<insert line that triggered decision\>". I do like my growing workflow with cursor for troubleshooting. I'm still a bit unclear on how AI agents work/what I'm using but I do enjoy that more active rubber ducky situation. I'm hoping over time building out this app becomes easier and my workflow evolves a bit to have deeper thoughts past my surface likes and dislikes.