My hunch was that making this quicksilver-iron-gold intermediary product should be pretty feasible at 3/2P (5 cycle tape loop and spending 10 cycles to make 3 product), and converting it into the final output would be very simple. i'm sure there's more minor optimizations here i missed but i'm happy enough with just reaching sub-100.

i experimented with a qs-i-qs-s intermediary instead but didnt seem like it would go under 100. idk maybe there's more solution space in that direction