She asks "Is this true?" and proceeds to show that it is [How many mutations?]. She assumes that the human mutation rate is 1.2 × 10-8 per sit per generation. Multiply this by 7.16 billion people on the planet and you get an average of 86 mutations at every single base pair in the human genome.1
Many of these mutations will be deleterious and they will be quickly eliminated from the population if they are lethal or cause severe problems. Some moderately and slightly deleterious mutations will be present in the population because they haven't yet been eliminated by negative selection. (Some will have no effect if they are present in only one copy of your diploid genome.)
To a first approximation, the statement is pretty accurate. If it's true that most of our genome is junk then the nucleotide sequence is not important.2 As we sequence more and more genomes we should see heterogeneity at 90% of the base pairs in the genome. We haven't reached this sort of coverage but all available evidence is consistent with the idea that most positions can be variable.
1 I prefer a larger mutation rate of 100 new mutations per generation for a total of 112 mutations at every site.
2. This doesn't rule out functions that are not sequence-specific. Such functions are known to exist but there are no reasonable hypotheses that justify such functions for most of the genome.