Meld is certainly interesting; manufacturing is going to be a challenge for them with a lot of moving parts, but in any case, it may work for some soups/stews types of dishes, but it won't be compatible with most of Pantelligent recipes (asparagus, eggs, crepes, pancakes, bacon). And a Meld-style internal temperature probe alone would be significantly worse than Pantelligent for foods like steak and salmon; I know that at first glance it seems like internal temperature is the temperature to measure and optimize for, but you also have to control the external/surface temperature or else it's "unobservable" and will lead to either undercooked or overcooked meats! You can not simply do feedback around the internal temperature. (That's one of the key engineering insights of Pantelligent.)
I don't see how you can possibly just do feedback on the external temperature. How could you possibly guess if the meat is cooked through just based on the surface temperature? You really can't unless you make some assumptions on initial temperature and thickness at minimum.
You are correct. Initial internal temperature and geometry of the food are key inputs. Once you know these, simulation is possible, but it takes some work.
A few years ago I wrote an app called Sous Vide Dash that was initially targeted at professional level kitchens-and later white labeled by PolyScience, the leading provider of professional sous vide equipment. We did full thermodynamic simulation in that app, and modeled both temperature and pasteurization. To my knowledge it's still the only cooking app that has an embedded numerical solver for partial differential equations. In 2012 we won the StarChefs International Chefs Congress Innovator award for our work.
Popping back to the top of the stack, it's a very exciting time for precision cooking. As one of the founders of Meld, I can tell you there is a lot more to come.
This is one reason we chose the form factor that we did for Cinder, because we can automatically measure thickness. We found that's something rather hard for people to do, no one really wants to hold up raw meat to a ruler app on their phone, and yet is crucial to time estimation. Then we ask you what we don't know, so in the app you'd say: Chicken breast. We've also done some cool things to know when it's done, like how a rice cooker clicks off.
To me, it looks the most compelling because it's both the cheapest and most versatile.
I backed both Pantelligent & Meld. But if Meld would have kickstarted first, I wouldn't have backed Pantelligent.