notes.txt

- k2 and k3 don't "just work" (as in slides); we MUST set k1=2, k2=3 or change the circuit reduction (see compiler/utils.py)
- in round_2 we are expected to compute an extra value that gets ***removed*** in an assertion (!); this explains why paper computes sum up to n-1 but we must go up to n.
- poly.to_coset_extended_lagrange hard-codes 4
- "Expand shifted Z(ω) into coset extended Lagrange basis" -- actually Z(Xω)