A sample size of 1 is acceptable for general analysis and comparative quality assessment of a manufactured LOT among uniformly control-manufactured tools. There are no "statistical results" in n=1 studies, but for this sort of "testing", there needn't be. The single representatives were chosen blindly from an (assumedly) uniformly manufactured lot (i.e. negligible sampling variability), the manufacturer also blind (i.e. did not know which knife or any at all would be "tested"). We must trust the technician that he did not attempt to deliberately skew the results by slight variance in handling (angle of incidence, etc.), but such variance probably would have minimal effect anyway given the tools used and the technique observed in the videos. THUS, from this SINGLE "test" on SINGLE individuals, it is scientifically valid to infer that the same results would occur for ANY selected individuals in the lots, especially if the same technician were employed.
N=1 inference tests are standard in QC analyses even in the biomedical field in which I work. We test ONE sample out of a lot - only if the values we collect regarding the quality of that sample are below acceptable levels do we test a second sample, and only if the 2nd value-set does not conform with the 1st do we test a third sample. One of the principles of good scientific (and business) practice is to not waste product & resources unnecessarily...