Very good sentences
Rather than assume current changes in M affect future AD with long and variable lags, I assume current changes in the expected future path of M affect current AD, with almost no lag at all.
That is Scott Sumner, pointed out by Arnold Kling.