Just to avoid any possible confusion in the way the note appears, in the sideline with 15...c6 white does not win a pawn with 16. Bf4, it is just a solid alternative to 16. Qxg7.
I apologize, as the note may appear misleading.
Peter Radomskyj vs. Mark Kernighan
Sort:
Aug 20, 2016
0
#3
It appears black could technically have lasted longer if he had tried 17...Kd8. I have yet to even find a forced checkmate after 17...Kd8, but I don't doubt one exists. But if 17...Kd8 after 18. Bg5+ no doubt white is winning. It looks like black's best would be 18...Qe7 (not 18...Be7 due to 19. Bxe7+ Ke8 20. Nxc7# or 19...Qxe7 20. Qxe7#), and that is obviously horrible for black.
annotations courtesy of Radomskyj