Find all sets of five Wordle words without repeated letters

solve.py

import glob
import numpy as np
from itertools import combinations
from time import time
t0 = time()

words = []
for file in glob.glob('wordle*'):
    with open(file) as f:
        for word in f.read().split():
            if len(set(word)) == 5:
                mask = sum(2**(ord(c)-97) for c in word)
                words.append((word, mask))

maskss = [np.array(list({m for _, m in words}), np.int32)]
for _ in range(4):
    maskss.append(np.unique(np.hstack([
        maskss[-1][maskss[-1] & m == 0] | m
        for m in maskss[0]
    ])))
    print(len(maskss[-1]), time()-t0)

winners = []
for (word, m) in words:
    for M in maskss[4]:
        if m & M == m and M - m in maskss[3]:
            winners.append(word)
            break
print(len(winners), 'winners', time()-t0)

for c in combinations(sorted(winners), 5):
    if len(set(''.join(c))) == 25:
        print('-'.join(c).upper())
    
print(time() - t0)

I use 26-bit ints to represent alphabet subsets that can be built with the Wordle words. I start with the 5-sets from the Wordle words, then compute 10-sets, 15-sets, 20-sets and 25-sets. Then I use that data to find the words that are part of a 25-set. There are only 28 such winner words, and 28C5 is small enough to recreate the actual sets of words by brute force, so I do that. To run this, add the two Wordle list files Matt links to in the video description.

Example output:

547117 0.6589927673339844
1020296 5.930372714996338
36183 9.305698871612549
3 9.440155267715454
28 winners 10.034610509872437
BEMIX-CLUNK-GRYPT-VOZHD-WAQFS
BLING-JUMPY-TRECK-VOZHD-WAQFS
BLUNK-CIMEX-GRYPT-VOZHD-WAQFS
BRICK-GLENT-JUMPY-VOZHD-WAQFS
BRUNG-CYLIX-KEMPT-VOZHD-WAQFS
BRUNG-KEMPT-VOZHD-WAQFS-XYLIC
CHUNK-FJORD-GYMPS-VIBEX-WALTZ
CLIPT-JUMBY-KRENG-VOZHD-WAQFS
FJORD-GUCKS-NYMPH-VIBEX-WALTZ
GLENT-JUMBY-PRICK-VOZHD-WAQFS
JUMBY-PLING-TRECK-VOZHD-WAQFS
10.190529823303223

Before the 11 solution sets, you see some statistics, like there are 547117 different 10-sets of letters that can be created using two Wordle words. The floating point numbers are the time in seconds from the start of the program. This is from a run on replit.com, couldn't run this on a better computer yet (but I do have a paid account there, so it's not that slow).