Ouch... Wild !
A computer would still be good, but here is what I can see.
2.Nc3 (?) is unambitous. Without the c4 push, you cannot exerce serious pressure against Black's center.
5.e4 was best answered by 5...c6 holding the center.
6...Nc6 (?) : 6...Bf5 with a probably good Caro-Kahn-ish position (compare with for instance 1.e4 c6 2.d4 d5 3.Nc3 dxe4 4.Nxe4 Nf6 5.Nxf6 gxf6 which is book). Here d5 is going to be a problem sooner or later.
9.c4 ?? loses a piece, when f3 holds everything, though with a better game for Black after 9.f3 0-0-0.
11.Qf3 loses less material, but it is just a piece down after 11...Qxf3 12.Nxf3 0-0-0, so I prefer the text.
14.Rd1 : I have the firm conviction that White should escape the perpetual or mate after 14.Qxa7, and thus should be winning. Something crazy like 14.Qxa7 Qg3+ 15.Ke2 Bd3+ 16.Kd2 but this is such a mess you should rather trust a computer.
23.Ke2 ! looks winning (again, check needed - pun intended)
In the 23...Qe3+ variant, you miss the shorter 26...Qxb4#.
This is a neat game I played a few days ago. I don't have a computer to analyze it with, but my opponent and I did spend a good 2 hours analyzing and recreating the game in the post mortem (it wasn't a casual game, so we didn't notate until after the game was finished, because it was such a crazy game). Would anyone mind telling me how me or my opponent could have improved? Thanks!