same process, same near heart attack - especially some days ago I had malicious activity on my platform and thought that now they actually found something.
Apparently you can trigger an Action (e.g. prevent uploads) when the billing alert triggers, but then my platform wouldn't work anymore, just because AWS had an issue. Also insane that Amazon still hasn't send an email to clarify.
Same - just had some malicious bots running through my platform last week and really thought they found a security hole after all. Even though the amount sounded ridicoulus, I got quite nervous and a very bad feeling when I logged-in AWS and saw that price.
I have built something similar for kids once, also using tensorflow: a game where you control a goal keeper with your body. Optionally kids (ok and adults) can color the goal keeper(https://dibulo.com/world/frutball
Tensorflow.js made it easy and it was quite fun to build. I still use it for breaks, where I want to move a bit and have fun
Back in the late eighties there was a IIRC NewTek video toaster for the Commodore Amiga which was a video / genlock system and there was a kind of interactive demo where you try to hit a ball / play goal using the camera. It's cool to see this type of idea again.
I would try using AI to book flights - then double check if I can't get a better offer. Do this a couple of times and when I see AI is as good or even better at getting me flights, then sure, why not use it.
Extrapolating from my experience testing it for coding tasks the result is not reliable even if it was right a couple of times. A risk I'm not willing to take. And I can't say that AI powered chat assistants on web pages have been much help either.
You can even automate this kind of testing in the AI model. I think the Google ADK has a built-in system for tests you use to confirm the reply quality.