One of the comments to Stephen Wolfram's blog post on Ada Lovelace, mentioned that WolframAlpha (verified in Mathematica as well) provided slightly different results for the question of Ada's age when her father, Lord Byron, died, if the question was asked in two different ways.
Using "death of lord byron - birth of ada lovelace" gives 8 years, 4 months 22 days, while "age of ada lovelace when Lord Byron died", gives 8 years, 4 months, 9 days.
On expanding the details in both answers, the resulting dates used to calculate the answer are the same. Where does this slight difference come from?
That's not true. Both inputs give 8 years, 4 months, 9 days.
Well, it looks like someone at Wolfram fixed the problem. I was there at the time of the blog post and a few days later when I looked at the two cases. Makes you wonder if you can really trust the answers you get, though.