And to cook foods without overcooking requires low power settings (~20%) for long periods of time (6-15 min), but constantly checking the temperature, guesstimating another x minutes, rinse and repeat... there has to be a better way!

It seems like building a cheap low-res IR camera in the roof of the microwave would be such an obvious idea. I punch in my desired 130°F, and it alternates between heating and waiting -- as soon as it detects hot spots or any food that's reached the target temperature, it backs off, then starts up again until the whole item is at desired temperature. Your food is never overcooked, and never undercooked. You could even defrost perfectly, and/or hold a temperature for really large items to penetrate the inside.

Yet there seems to be vanishingly little progress. Searching online yields 2015 patent that seemed to go nowhere [1], a dismissive article from 2018 [2] about a hard-to-use IR accessory, and a deep learning paper from last year [3].

So I thought I'd ask here, since there are bound to be engineers with experience. Is there something I'm missing here that either makes it impossible to build (e.g. microwaves would fry the camera electronics or something) or somehow so inaccurate it wouldn't be worth it?

I don't know if this is a startup idea, or just something I wish the big brands would start producing...

[1] https://gizmodo.com/infrared-microwave-lets-you-watch-your-food-change-colo-1685143405

[2] https://www.theverge.com/2018/3/19/17139114/cmicro-microwave-temperature-sensor-system-kickstarter

[3] https://www.mdpi.com/2571-5577/3/1/13/htm