I built a super cute “Rate My Matcha” app in one night, basically as a little cozy devlog experiment to see what I could actually ship after work. I started around 7pm and gave myself four hours, because I wanted most of the features done before my favorite cafe closed. I used Lovable to scaffold the web app (camera input + title), then I jumped into the generated GitHub repo in VS Code to tweak the parts that needed more control—like fixing the camera and improving the rating logic. For the matcha rating, I first thought I could just use RGB, but lighting changes everything, so I switched to HSV (hue/saturation/value) for a more accurate color score. After that, I made the whole thing look cozy with pixel art (Procreate on my iPad) and a Figma mock, then passed my mockup into Lovable to transform the UI. I connected Supabase for storing matcha photos + reviews and added Google auth so everyone can save their own personal gallery. At the end, I took the app to a real cafe and tested it on an actual cup of matcha. It’s not perfect yet, but that’s kind of the fun part—I want to keep collecting labeled matcha photos and eventually train a little ML model. The big takeaway: AI tools like Lovable can seriously compress a week of setup/styling into a night, and you get better at prompting + debugging “someone else’s code,” which is honestly a real work skill.