If p > 0.05, then the tested value does not differ significantly from the group. In other words, there is no reason to assume that the sample was taken from a different population. In this case you keep the null-hypothesis, i.e. the sample value does not deviate from the overall group [average/distribution].

I think you are confusing yourself with the assumption always made about the null hypothesis. The null hypothesis here is the mean is 18. Just run the test and if Pvalue is less than alpha, reject the claim or null hypothesis, if p value is greater than alpha as in the video, fail to reject the null. Most often in linear regression, the assumption for the null is, there is no linear relationship between response and explanatory variable (SLM), when you actually suspect there should be from a quick boxplot or a quick regular plot. We end up with a Pvalue less than alpha level and reject the null. the assumption is different in this case using the mean. The null simply is saying the length of 18 is the true length. I hope this clarifies your confusion.