How are we coding Android apps with AI in Feb '26

My little journey so far:

My web colleagues have been embedded in Copilot with VS Code since it's inception while I was tinkering around with ChatGPT. I am one of only 3 android devs. I've been using ChatGPT and Claude but almost always via the browser. I've tried Gemini multiple times for semi complex issues and found it failing hard and fast so I am disinclined to use Gemini again for the forseeable future. I like many others am gravitating towards Claude's Opus model. I spent two days recently talking in circles with ChatGPT for a complex architectural unentanglement and Claude came up with good answers in a couple of hours. However Claude is extremely limited with it's quota. My web colleagues are using the basic Claude plan then swapping to Copilot to get them through the day.

I've dabbled in the CLI tools but found them making changes I found difficult to trust and read through properly, and preferred discussing solutions first to refine them.

With agentic AI we have another milestone, I think this year people will completely stop writing code themselves. I see models can be changed in Android Studio now, I am going to have to start using "code-first" AI solutions for my day-to-day Android development.

So what is your story with AI and Android:

• How are you developing day-to-day, are the Android Studio tools sufficient?
• What models do you use
• What tools do you use - Junie, Cursor, straight CLI etc.
• Do you build entire modules with tests?
• How do you keep costs down?