And here's the imaginary part:

@dharr Okay, I understand, and I withdraw my comment about it being incorrect. And I put a more-formal withdrawal at the top of the Reply.

[Edit: I understand dharr's followup explanation, and I withdraw my comment about the work being incorrect. I didn't realize that it was just solving (correctly) a different problem than the OP intended, specifically one where certain edges are forbidden.]

@dharr Your algorithm for reducing the Hamiltonian cycle to a Hamiltonian path is incorrect. In particular, I don't understand how you chose your edges. If I apply my dummy-vertex algorithm (from my Answer below) to your points, I get a significantly shorter minimal path, 7.3 vs. 9.8. You can easily verify from the plots that my path is Hamiltonian and my edge weights are the same as yours.

restart:
pts:= [[0,0], [1,2], [4,2], [3,1], [5,1]]:  
n:= nops(pts):
GT:= GraphTheory: LA:= LinearAlgebra:
G:= (GT:-Graph@Matrix)(
   n+1, 
   (i,j)-> `if`(i=j, 0, `if`(i>n or j>n, 1, LA:-Norm(<pts[i]-pts[j]>, 2))), 
   shape= symmetric, datatype= hfloat
):
(MinDist, HamPath):= GT:-TravelingSalesman(G, startvertex= n+1):
MinDist-= 2; HamPath:= HamPath[2..-2];
                  MinDist := 7.30056307974577
                   HamPath := [1, 2, 4, 3, 5]
#Plot:
H:= GT:-InducedSubgraph(G, [$1..n]):  #Discard dummy
GT:-SetVertexPositions(H, pts);
GT:-HighlightTrail(H, HamPath);
GT:-DrawGraph(H, axes= frame, scaling= constrained);

I think you're posting in the wrong forum. As far as I know, no Meplesoft product has a subpart named "Workday".

I wonder how ChatGPT came up with the non-existent command Optimization:-LinearSumAssignment. It had to have read that somewhere; it doesn't just make up fake Maple command names on a whim.

@C_R The command that you're thinking of is called unames.

@Scot Gould I've also noticed a speed improvement the past day or two 

@nm The counter can be reset, without using restart, by

`tools/genglobal`[1](_C, 1, reset);

That form of the command just does the reset without returning a name. So, the next call to `tools/genglobal`(_C) will return _C1.

@Scot Gould I've experienced the same thing as you: The norification flag used to do something (not something that I find useful---but that's beside the point) and now it just says Loading.

See the help page ?simplify,size.

@mmcdara Those two things are not necessary for your example, but might be needed for some generalization that you have in mind. What I can't figure out is why assigning invfunc[phi] doesn't require the unprotect, but assigning both invfunc[phi] and invfunc[invphi] does require it.

The use of invfunc can be implemented like this:

restart:
phi:= u-> (u^(-theta)-1)/theta:
invphi:= unapply(solve(phi(x)=u, x), u):
unprotect(invfunc): invfunc[phi]:= invphi: invfunc[invphi]:= phi: protect(invfunc):
C:= (u,v)-> simplify((phi@@(-1))(phi(u)+phi(v))):
C(u,v);
            

To understand better what's going on here, check out showstat(`@@`).

@janhardo I don't find it at all surprising that ChatGPT can give you an extensive discussion of the infinitude of primes, which is one of the most-ancient topics of formal abstract mathematics. But how does it involve Maple or any computer language?

Yes, I have noticed that most operations on MaplePrimes have been much slower for a few weeks.

I deleted your Question from today because it is essentially the same as this one. If you have additional information, put it in this thread.