Coding - me & Claude pt 2

Users, eh, got to love them.

I created at app to explore agentic coding, help structure my golf hobby (obsession) and to learn more about React Native - see previous post. I then released it to the world and have received unsolicited feedback in return.

Responding to feedback exposed capabilities gaps in my workflow. In my workflow, I had fallen into the oldest trap of all:

• Claude (mostly) amending the code

• I would review the code and suggest improvements (repeat step 1)

• When complete, I would review the output via the UI.

Basically I had become (if I were being paid like normal) an expensive manual user tester.

I have spent my career replacing manual testing with automation in workflows enabling businesses and government departments, including departments delivering Critical National Infrastructure, to reliably release changes quickly. And here I was doing slow, unreliable and repetitive manual testing in my own time.

Automation had to be the key. I had already installed maestro to capture screenshots for App Store and Google Play listing so I was part way to a solution.

With my thinking and Claude's execution, my workflow is significantly improved and automated. A pre-push hook captures the latest version of the UI, compares each screenshot to a baseline (with a small tolerance margin) and rejects the push if this is any variance.

Command line arguments allow me to skip this step (in the interest of time because it is slow - about 20 seconds) or update the baseline if the UI change was intentional.

I can make changes based on user feedback with confidence that the changes have not cascaded through the app with unintentional UI amendments.

I maybe on a journey to actually loving feedback from my users...

⛳ 🏌‍♀️ 🎉 #ClaudeCode #RN #AppDevelopment #LoveYourUsers