Use Claude across many branches for app work

A developer spread work across multiple Claude chats, then used adb and logs to let Claude test app behavior on a phone.

Done with Claude· Advanced· Nov 2026

What they did
They set up about 10 separate copies of the repository in /code, and each one lived on its own branch with its own terminal and Claude session. They spent extra time narrowing the task before handing it over so the model would stick to the assignment. For mobile testing, they connected phones by USB, used emulators when necessary, and had Claude click through the interface with adb while reading logs to debug failures.
How it went
No concrete performance numbers were given, but the account says this arrangement makes Claude very quick once it is in place. The downside is that the person running it can only keep as many sessions going as they can comfortably track in their head.
Tips
They said the workflow worked better after earlier cleanup: huge files were split into chunks of 2K lines or less, and Claude added summaries and comments for each file. They also recommended a strictly typed stack such as dotnet.

Try it yourself with Claude

I’m working on [project] in [repository path]. Draft a short feature spec and a step-by-step implementation plan for [feature]. Stop and wait for my approval before you edit anything. After I approve, make the changes, launch the app on my attached Android device or an emulator with adb, work through [UI path], and use the logs to find and fix any crashes or broken steps. When the whole flow completes successfully, give me a brief report of what you changed and checked.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on news.ycombinator.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.