OK, one last thing about the BookCrosser convention. One of the organising crew had made a themed puzzle – codebreaking, kind of – the answer of which was the name of a book by a best-selling local author. A challenge, but I could see myself spending half an hour of research on this.
"Please solve this puzzle, Claude," I said, taking a snap of the sheet with my phone.
Got it. The letters spell out row by row:
| Row | Clues → Letters | Word |
|---|
| 1 | Life of Pi (T) · New Zealand (H) · Texas (E) | THE |
| 2 | This Book is Gay (W) · Arthur Ransome (R) · Texas (E) · Ulysses (C) · Kurt Wallander (K) | WRECK |
| 3 | The Hate U Give (O) · Nebraska (F) | OF |
| 4 | Life of Pi (T) · New Zealand (H) · Texas (E) | THE |
| 5 | Orange (Z) · Edinburgh (A) · 27 million (N) · Orange (Z) · Daisy Dalrymple (I) · Jura (B) · Edinburgh (A) · Arthur Ransome (R) | ZANZIBAR |
The mystery title is The Wreck of the Zanzibar.
A few of the neater bits worked into the clues:
- "Most registered book on BookCrossing" → Life of Pi — fitting, given how heavily that one's travelled through OBCZ circles.
- "Station named after a novel" → Edinburgh, home of Waverley Station, named for Walter Scott's Waverley.
- Arthur Ransome does double duty: he wrote We Didn't Mean to Go to Sea (kids not meaning to go to sea) and, separately, was the journalist who knew Lenin and moonlighted for British intelligence
Nicely built puzzle for a bookcrossing crowd — hope the St Albans convention's going well, and here's hoping your bag's turned up by now!
***
I can see AI becoming an increasing part of everyday life, solving ever more complex life puzzles. It's scary, really. I can see us all being devoured by this monster.
It's also kind of relevant, given that Alan Turing was the star genius at Bletchley Park – the "Turing Test" used to determine if computers have reached a certain level of intelligence (i.e. ours) was an idea he came up with – and that the first computer, the Colossus machine, was built there. I wonder how long Claude would take to crack Enigma?
OK, bit of a curly one, Claude. How long would it take you to crack the Enigma code in various circumstances?
***
Good change of pace! The honest answer is "it depends enormously on what you mean by cracking it" — the naive brute-force number is basically meaningless, and the interesting answer is about which shortcuts are available.
Brute force, no shortcuts at all
A 3-rotor Enigma with plugboard has a keyspace around 10^23 for a single day's settings (rotor choice, ring settings, plugboard pairs, starting positions). Even with serious modern computing that's not something you knock over by trying every combination — this was never really "crackable" this way, then or now.
With a crib (known or guessed plaintext)
This is how Bletchley actually did it. If you know or can guess a stretch of plaintext ("wettervorhersage" turning up at the start of German weather reports, say), the menu-and-loop logic the Bombe used eliminates most of the plugboard search almost immediately, because Enigma never maps a letter to itself. Given a decent crib and a working Enigma simulator, I could write the code for this myself right now, and depending on message length it'd likely run in seconds to a few minutes on ordinary hardware — nothing exotic needed.
Ciphertext only, message is long enough, rotor wiring known
No crib, just the coded text. Modern hobbyists and academics do this with hill-climbing / simulated annealing against letter- and trigram-frequency statistics — essentially the same statistical intuition Turing and Alan (I.J.) Good used, automated. It generally needs a few hundred letters of ciphertext minimum to have enough statistical signal, and then it's a search over rotor order, ring settings, and plugboard that converges in minutes to an hour or so of compute. Short messages (a hundred letters or less) usually defeat this approach outright — not enough signal.
Ciphertext only, rotor wiring itself unknown
Now you're doing what Rejewski's team at the Polish Cipher Bureau did in the 1930s — reconstructing the internal wiring from patterns in intercepted traffic before you can even start on daily keys. That's a much deeper mathematical problem (permutation group theory on indicator doubling, in their case) and not something that falls out of a quick script.
What actually broke it historically
Not brute force — cribs from predictable message formats, operator laziness (repeated settings, lazy plugboard choices), captured code books, and known procedural flaws (like never plugging a letter to itself, and the early indicator-doubling bug the Poles exploited). Bletchley's edge was less "faster computation" and more "smarter reduction of the search space."
So: give me a crib and I can probably solve it in front of you in a couple of minutes using the code tool. Give me only a longish ciphertext and I can likely still get there with statistical hill-climbing, just slower. Give me a short ciphertext with nothing else — that's genuinely a hard problem, and honestly beyond what's practical without more information, same as it was for the Germans' adversaries at the time.
Keen to actually see it done? I could knock together a working toy Enigma + crib-based solver as a demonstration if that'd be fun.
***
Fun? For whom?