[seqfan] Re: : The alternating sum in A196020

[William Keith]

Seems plausible to me.  I would suggest something along the following
lines: you are including or excluding divisors of size k by adding the k-th
column.  The k-th column becomes "necessary" when you must add or cancel
divisors of sizes 1, 2, ... and k, when you have only had to add or cancel
at most k-1 of them before, ergo it appears at n=k(k+1)/2.

(Conjecture checks out to n=100 000 easily, btw.)

William Keith