You can't do straight averages of salaries because they're heavily skewed, especially in the case of tech. Either use average + confidence interval or use medians instead. (relatedly, don't use area charts for single statistics)
Cool, very interesting but is this data just scraped from AngelList? If so, is there a risk that its more representative of AngelList's growth as a platform / relevancy amongst employers?
That's the second problem with the dataset. (sampling bias for AngelList's preference toward startups)