Something to steal
Paste this into your AI assistant before you start troubleshooting:
Before you suggest any fix, list the likely causes, most likely first. Then give me ONE change to try. Wait for my result before you give me the next one. Never give me a list of changes to make together. After each test, tell me what the result proves and what it rules out.
Before and after
Phone alerts that never arrived. My business dashboard sends an alert to my phone when something urgent needs me. It worked while the browser was open in the background. When I swiped the browser away, alerts sat unseen until I opened it again.
- Before: the AI asked me to change four phone settings at once. I refused. If all four went on and the alerts started working, we would never know which one did it - and I would be stuck with three settings I didn't need.
- After: one change only - "Put unused apps to sleep" turned off. Test alert sent with the browser closed and the phone locked. It arrived in about a minute and a half. The same test before the change: nothing until I opened the browser. One change, a clear before and after, done.
Images that wouldn't show in my notes. The AI kept guessing - the link format, then the search index, then the list layout, then the screen - and edited files as it went. One note got damaged along the way.
- Before: a new theory every turn, a new edit every turn, no progress.
- After: I gave it one note that worked. It put the same image into six different layouts and changed only one thing in each. The difference showed up in a few tries, and the fix was obvious.
Why this works
Each test answers one question. A bundle of changes answers none - it only tells you that something in the bundle mattered. One change at a time feels slower, but it is the fastest path to a fix you can trust and explain.