Our first approach to Rate built the product in 2 halves.  The process for that was straightforward, but had two weaknesses. First, it forced most of the machine into Period 4, which heavily constrained the number of manipulations each arm could do and thereby constrained the use of space.  Second, doing the triplex bonds at the end involved rotating or pivoting the whole molecule, which was quite area-costly.  The best we got with that strategy was 88A.

By contrast, doing triplex first was never larger than 82A--clearly better.   Our opening bid optimized down to 74, and the first version of our redesign was 64. A few tweaks more yielded the machine you see here.

=========

In our view, the main challenge was "swapping sides" with one salt and two metals without crashing into the growing product. You can see here that the first copper-QS component needs to wait a few cycles before it can be swung into position, reducing free latency and constraining input positions.