I thought he was joking in that first thread, there are so many mistakes in it
including the 30 move draw.

But if you set the computer as Black with the lone King to refuse
a draw then it will not allow a three fold rep so walk into the one of the
very rare and unforced positions where two Knights can actually give a mate.

Will a box do that?
If so then Moonbus is right with:

"Checkmate with K+2N vs K is possible, though not forced (for humans within tourney limits)"

I've posted in the past and very recently that three Knights cannot mate a lone King.
Of course they can, last time I did it I left it unedited just for the hell of it.
I mis-read it somewhere.

Originally posted by moonbus
I vaguely recall an article in an IT journal, not a chess journal, some years ago, claiming that the K+2N vs K problem had been solved. Alas, I can find no reference to this on the Internet. As I recall, the program ran for several days (well over tournament time limit) and the mate required many more moves than the 50-move draw-limit, of which all but the l ...[text shortened]... nd made mistakes (drawing in won positions), we mere mortals may accept that K+2N vs K is drawn.

Originally posted by Paul Leggett
The king and two knights vs king ending was solved several years back as part of the 4 piece endgame tablebase. The ending was known to be drawn as part of theory, and the tablebase confirmed it.

Perhaps there was another element involved in the article- memory is always a tricky thing. My wife reminds me all the time!