Out of curiosity, I switched the meter linked above to S type thermocouple, so it thinks the Seebeck coefficient is 7 times smaller than K type that is actually plugged into it. The temperature readings are wrong, of course, but the fluctuations are about 0.2F . So, in principle its capable of reading about 0.01 C. Maybe with some hacking it could display another digit if one made a Seebeck lookup table 10 times smaller than K type.