Repeated failed troubleshooting loop wasted significant time/tokens (Expo Go tunnel setup)

During a long session, I asked Claude Code to help get a React Native/Expo iOS app (previously stuck on a local-network connectivity bug) running via Expo Go tunnel mode.

What happened:

• Claude gave multiple incorrect, unverified instructions about where to find UI elements (e.g. "Enter URL manually") in the current Expo Go app, which turned out not to exist / not match the actual app UI.
• A QR code image was generated and reportedly sent multiple times, but never actually appeared on my end.
• This repeated back-and-forth (wrong instructions, then a fix, then still not working) happened multiple times in a row without resolution, consuming a large amount of conversation/tokens with no working result.

This is part of a broader pattern I've experienced across this session (and, in my view, prior sessions) where the assistant states something is fixed/verified/sent, but it isn't actually working on my end, and I have to spend more turns discovering that. I understand the assistant can't see my physical device, but stating unverified UI details as fact (rather than asking me to check first) is what caused the repeated failed loop here.

Filing this as feedback per the in-app guidance ("report the issue at https://github.com/anthropics/claude-code/issues").