This has come up in a few places recently. I have been revisiting GM Andrew Soltis's Studying Chess Made Easy, to refresh my memory of the various exercises he recommends for studying. I have also been reading GM Jonathan Rowson's Chess For Zebras at bedtime. Both them make compelling arguments that people make the mistake of trying to absorb vast amounts of information when learning chess, but they should be focusing on skill instead of knowledge or as Rowson put it in another way, know-how vs know-what. To that end I have begun focusing more on practical exercises, what that translates into is spending a majority of time playing games, and analyzing positions.
I also like Aagaard's and Rowson's suggestion to view a game of chess as a series of problems to be solved. Your goal is to successfully solve the problems that appear before you on the board.
Taking all this together along with Aagard's suggestion of starting doing 20 minutes of work a day for a few months and then building that up to build up your endurance for work without burning out, I am spending my 20 minutes analyzing a position each session from Israel Gelfer's Positional Chess Handbook. Once I am done I look at what is in the book, and then I also load the position into Fritz and score both my line and the book line using the Calculation trainer. Tonight I used a small analysis board to set up the problem, but I might alternate that with a tournament set, or just loading the position in Fritz, starting the calculation trainer, and going from there.
Here is the position I looked at tonight
The book line was f:e5 B:e5 B:e5 Q:e5 d4 Qc7 c:b5 a:b5 d:c5 Q:c5 Nd4.
I had stopped looking at Q:e5 and thought it wasn't promising for white, forgetting my rule that I was going to try and look 5 ply deep.
Instead I had decided on c:b5 a:b5 a4 b:a4 b:a4 creating an outside passed pawn. Fritz thought the line was ok, but the idea played in the game of limiting the scope of the black bishop and posting the N on d4 is probably a better plan. It considers c:b5 the best move, so maybe the thing to do is try some monte carlo simulation or running an engine tournament and see what their results favor.