How RC2's π table was made

Author here. After the NIST curve seed mystery was solved earlier this week, I set out to find other unexplained mystery constants in old ciphers, and had Claude Code (Opus 5.5) systematically scan for candidates and attempt to reverse-engineer their construction. One of these was RC2's 256-byte permutation table, created by Ron Rivest. It was described as having been derived from the digits of π - but exactly how seems to have remained a mystery.

Claude attempted several different approaches and eventually, drawing on a historical clue in how the MD2's substitution table was constructed, found a surprisingly simple method that reproduces all 256 entries exactly. I've shared the derivation and reproducible code. I couldn't find any evidence of an existing public answer to this, but will certainly update if one comes to light.

Now I wonder when will people find how Carmack found 0x5F3759DF