I'm curious - have you tested with oils with various smoke points? How does your device know which smoke point to monitor for? That's a 300F range (225 for Sunflower oil - which nobody should cook with, but if you want to prevent fires, look for the people doing things they shouldn't do, but that are convenient. 300F for butter, 520F for avocado oil)
Also, since I have a thing for flambee dishes - how does it deal with pyromanic chefs? ;)
It doesn't know which smoke point to monitor for. However, the aim (at least for the MVP) is for it to work the majority of the time for the majority of users - and critically to alert several minutes before there is any flame. So while sunflower oil may start smoking a small amount, it would be several minutes more getting to flame, by which time the alarm should have sounded.
If you put a frying pan on with oil in and leave it, the device will alert you at around 450F. This is a fixed threshold (along with some other checks, e.g. we also consider the rate of change of temperature). This gives reasonable behaviour in most cases, and ensures you don't get nuisance alarms when frying on high heat. In cases where individuals have an unusual cooking pattern which causes unreasonable false alarms we can adjust the model parameters to increase/decrease sensitivity in particular situations. In the future we may be able to adjust these automatically but for now there is not a great deal of automated "learning" involved.
If you're a pyromaniac chef, I think you would get some alarms from that! Again this is a bit of an edge case so we haven't but much effort into that for the first iteration of the product, definitely something for the future though - I can imagine it being a challenge to distinguish flambee and a real fire though, since flambee is real fire!
Also, since I have a thing for flambee dishes - how does it deal with pyromanic chefs? ;)