Skip any paragraph if it describes tech that a previous solution has demonstrated.

First I made a machine that compared an exo sample atom to a match pattern atom every 6 cycles.  It used 6 quint glyphs, carrying the two atoms it was comparing along with two other atoms that duplicated into the 6 different combinations.  This ended up with a wsum of about 1360, and with some math and estimation I figured out that to beat it a machine would need to compare atoms at most every 15 cycles.  A 12-loop or a 9-loop seemed like the best candidates.  I kept trying to figure out how to get hexarms next to the quint glyphs to work, but with only one atomic salt input couldn't figure out how to make that work efficiently.

So, I went back to trackloops, and used the method of carrying the 4 cardinals alongside the 2 atoms being compared.  That's what you see here.  I test 3 combinations on the left quint, and the remaining 3 on the right.  To see if the atoms match, I check if a cardinal atom is left at the end.  I dispose of the quintessence by having the arms that previously held cardinals carry them to the output.

The 6-long exo samples are perfect for my wasteless input strategy.  I need to make 12 comparisons per exo sample, so I use all 6 atoms from the sample itself and 6 more from the match pattern input.  This does mean that I need to get 6 salt to the end for the end conditional and output, which isn't done the most efficiently.

I predict that my strategy could be improved to 1250 with layout changes.  I can also imagine someone managing a 10-loop with a similar approach, maybe involving two trackloops at the bottom.  I predict the winning solve will have a three-digit wsum.  I currently can't conceive of any helpful black magic (after all, you have to drop lots of atoms on quint glyphs) but expect to be astounded by the stream.