I think I generally agree that some form of IR is the future of temp logging. It's cheap, relatively simple, low latency, and highly consistent. All attributes that you want for temp logging.
I do think there are some valid concerns though. Beans aren't a uniform color, even within lots. How does this sensor deal with chaff while it's still on the bean but about to be shed? Some beans roast much darker than others as well. I'm curious what the Bullet R1's sensor aperture is, since all of this makes me think due to bean agitation/movement and the need for lower variation in color changes the IR sensor is observing an area larger than what one would traditionally envision with a thermocouple. Smoke can effect readings a decent amount as well but I think there are newer methods that don't have this limitation. Last is the emissivity adjustment, which is the easiest critique but probably the most valid. Jonas is correct, consistency is what matters most. But that's a double edged sword with IR sensors, since the reason consistency matters is because the derivative of the temperature curve is far more important than the actual temperature curve itself. If you adjust for one end of the spectrum or another you're manipulating the shape of that curve. When comparing curves between roasts on a single roaster that might not be a problem, but it definitely removes some of the reproducibility between R1s and almost completely removes it when comparing roast curves between different machines. I believe many modern IR devices allow for adjustments based on a curve or table, is this possible the R1's implementation? Last, IR sensors will fail more often than thermocouples and be more difficult to troubleshoot.
I do wish you guys at Aillio all the best though, i'll be looking to pick up one of the March shipment R1s.