Update: my home made plant watering robot dumped 2 gallons on my desk while I was at lunch
So about 6 months ago I finally got into that hobby everyone talks about, training a tiny AI model on my own photos to spot when my plants look thirsty. It ran great for weeks, honestly I was proud of it. Then last Tuesday I went to grab a sandwich in Decatur and came back to find the pump had kicked on and would not stop. My sensor got stuck reading "dry" and the little model just kept agreeing with it, watering my poor fern until the pot overflowed onto my keyboard and a stack of mail. I lost a $60 mechanical keyboard and my desk still smells like wet dirt. The fix was dumb simple too, I just added a rule that cuts the pump after 4 seconds no matter what the AI says. Back in the day I watered plants with a cup and my own eyeballs and somehow nothing ever drowned. Now I'm wondering if anyone else has had a model get way too confident about something that was clearly busted, and how you put a leash on it?
Reminds me of my brother's garage door opener that kept opening every time the neighbor's dog barked. He fixed it with a piece of tape over the sensor, so I guess you two went to the same school of engineering.