Appendix A — The Moves Index {#appA}

Fifty puzzles and sixteen olympiad tools compress to thirty-two working moves. One line each; the line is a reminder, not a lesson — the lesson lives at the reference. References read chapter·puzzle: 5·6 is Chapter 5, canon puzzle 6; a plain Ch 9 points to a chapter's worked material rather than a numbered puzzle; C·17 means Appendix C — the Reserve Rack — canon puzzle 17. This index is the book's real curriculum. The puzzles were the delivery system.

See it differently

• Switch the frame; question the boundary. When the facts are bizarre but consistent, the blocking constraint is usually one you added — 4·12, 4·13, C·17.
• Break functional fixedness. Ask what each object is, ignoring its label: a box is also a shelf — 4·14.
• Change the space. Add a dimension when the plane is the prison (4·15); unfold the space until the answer is a straight line (C·24).
• Model the story as structure. Name the vertex and the tale collapses to a known kernel; measure progress in state-space, not cargo — 1·1, Ch 11.

Find what cannot change

• Find the invariant. A conserved quantity refutes without search — colorings for tilings, parity for move-systems; or find the monovariant that only falls, and the process must stop — 6·20, 2·38, Ch 9.
• Conservation ends the argument. Endstate totals and global aggregates beat episode-by-episode tracking; when positions lie, rates may not — C·29, C·28, 4·16.
• Hold the balanced state. Identify the dead positions and return the game to them every turn; the opponent's freedom becomes irrelevant — Ch 9, C·43.

Treat knowledge as a substance

• Budget outcomes before designing tests. Count the possible answers, count each test's outcomes, and let the ratio dictate the design — 7·8, Ch 11.
• Assume, then break it — and use the guaranteed error. Hypothesize each case and let contradiction do the elimination; a certain lie is exactly as informative as a certain truth, and can be routed to cancel itself — 5·2, 5·3, 5·4.
• Declared ignorance is data. "I don't know," said aloud, filters worlds; public words mint common knowledge, which is more than everyone knowing — 5·6, 7·49, 7·10.
• Condition on the procedure. What a fact means depends on how it reached you; the sampling defines the odds — 6·33, 6·34.
• Knowledge can self-destruct. Statements about what you will know are not ordinary facts; feeding them into your reasoning can invalidate them — 7·50.

Compute a smarter way

• Anchor, then propagate. Plant the absolute facts, squeeze the tightest constraint first, and never guess while a forced move remains — 5·5, 6·41.
• Work backwards. The goal end of the tree is thinner; label the endgame and roll toward the present — 7·44, Ch 10.
• Recurse, then cache. Trust the smaller self; the art is choosing what a "state" must remember — C·31, Ch 10.
• Count another way. Complement (6·32), pairing and its exceptions (C·30), double census, expectation — one set, two ledgers, and the agreement is your theorem; an average of v forces some case ≥ v — Ch 11.
• Small cases propose; proofs dispose. Tabulate, conjecture, then prove — and when the inductive step leaks, strengthen the claim until it carries its own fuel — Ch 10, Ch 11.
• Prove greed with an exchange. Or spend the same ten minutes hunting its counterexample; unproven greed is a leap, not a plan — Ch 10.
• Binary-search the answer. When feasibility is monotone, optimize by deciding, log-many times — Ch 10.
• Interrogate the extreme witness. The largest, smallest, closest, or first element arrives pre-loaded with contradictions for free — Ch 9.
• Impose or steal a symmetry. Mirror the opponent and every future question is pre-answered; or show the extra move never hurts — Ch 9.
• Invent the pigeonhole boxes. The theorem is one line; the entire craft is manufacturing the categories — Ch 9.
• Declare WLOG, then design the casework. Quotient the symmetry first; choose the split so every branch lands on a finishing move — Ch 11.

Borrow, share, correlate

• Restructure the bill. Borrow a unit and return it (6·26); make the expensive items share (5·7). Temporary extensions reveal permanent structure.
• Engineer the joint outcome. Correlate fates so everyone lives or dies on one event (7·37); when shared state is tiny, design the protocol, not the deduction (7·9).
• Build commuting tools. Do, do, undo, undo — small operators that change little and preserve the rest are how you edit any system too complex to rebuild — C·39.

Audit before trusting

• Audit the invisible assumption. List your defaults in writing; attack the oldest first — 4·11.
• Distrust the surface. Verify the diagram, run the algebra against the gut, attack the statistics; whatever must preserve function must leak pattern — 6·22, C·21, C·40.
• Pin the random variable; stratify before comparing. Most paradoxes are notation crimes; most reversals are hiding in the case mix — C·36, 6·35.

Scale

• Decompose, estimate, multiply — then respect the exponent. Work in powers of ten, state bands not points, and never eyeball a doubling process — 6·46, C·47.

Other minds

• Model the maker. Go one level deeper than the crowd, then stop (C·45); artifacts built by intent are decoded fastest by modeling the intent (Ch 8) — and sometimes the surface is the instruction (C·42).
• Seek the fixed point. Make the description part of the described and demand consistency: x = f(x) — 6·48.

Appendix B — The Graveyard of Thinking Myths {#appB}

Every discipline keeps a trophy room; this book also keeps a graveyard. The six residents below were beautiful, plausible, and beloved. Several were founded on real experiments. All are dead. They are buried with respect and full honors, because each taught the field something by dying — and because a book that spends fifteen chapters teaching skepticism owes its readers the kill list. Visit; leave flowers; do not resurrect.

⚰️ Commercial brain training

The claim. Ten minutes of app puzzles a day raises memory, attention, and intelligence, the way curls raise a bicep. Why it spread. Real neuroplasticity science, an aging population happy to pay for insurance against decline, and honest-feeling receipts: scores on the trained games genuinely climb. What the evidence says. Owen et al. (Nature, 2010): 11,430 adults, six weeks — trained tasks improved sharply, transfer even to close cousins was approximately nil ⚖. Simons et al.'s 2016 review of the industry's evidence: the same. The FTC fined Lumosity's maker $2 million in 2016 for deceptive advertising. Practice is specific: the game trains the game. Verdict. ⚰️ Dead. Epitaph: HE GOT REALLY GOOD AT THE GAME. (Chapter 15's naming-and-recostuming discipline exists because this corpse is real.)

⚰️ Learning styles

The claim. Each person is a visual, auditory, or kinesthetic learner, and teaching matched to the style produces better learning. Why it spread. Preferences are real and easy to measure; the idea flatters individuality; an inventory-and-workshop industry grew around kind teachers who wanted to personalize. What the evidence says. Pashler, McDaniel, Rohrer, and Bjork (2008) audited the field: the claim requires a crossover interaction — visual learners learning more from visual teaching, auditory learners the reverse — and almost no study ran the right design; those that did found no meshing effect ⚖. Preferences exist; matching to them doesn't help. The content dictates the best medium: geometry wants pictures, for everyone. Verdict. ⚰️ Dead. Teach to the material's style, not the student's.

⚰️ Left-brained and right-brained people

The claim. Logical people run on the left hemisphere, creative people on the right. Why it spread. Sperry's split-brain research was real, spectacular, and Nobel-crowned (1981); personality quizzes and corporate workshops flattened a true fact about severed hemispheres into a false fact about accountants. What the evidence says. Functions lateralize; people don't. Nielsen et al. (2013) analyzed resting-state scans of 1,011 brains and found no individuals with a globally dominant hemisphere ⚖. Every intact brain runs both halves on essentially everything. Verdict. ⚰️ Dead. You are whole-brained; so is everyone you have ever argued with.

⚰️ We use only 10% of our brains

The claim. Nine-tenths of your brain idles, waiting to be unlocked. Why it spread. Often traced to a mangling of William James, who wrote (1907) that we use a fraction of our mental resources — a motivational claim about effort; the tidy "10%" was pinned on later and no one traceable ever measured it. It survives because it prints permission slips: the missing 90% is yours, for the price of the seminar. What the evidence says. Imaging shows nearly all of the brain active across an ordinary day; no region can be lesioned for free ⚖; and an organ that is 2% of body mass billing roughly 20% of the energy budget would be evolution's most overpaid idle employee ⚙. Verdict. ⚰️ Dead. You use all of it. Some days it doesn't feel that way, which is a different problem.

⚰️ The Mozart effect

The claim. Listening to Mozart raises IQ — so play sonatas to infants and they grow up smarter. Why it spread. Rauscher, Shaw, and Ky (Nature, 1993) found that 36 college students improved briefly on one spatial task after ten minutes of a sonata. The press deleted "36," "briefly," "one spatial task," and "college," and by 1998 the governor of Georgia was budgeting classical CDs for every newborn in the state. Note the chain: no infants appeared anywhere in the study. What the evidence says. Meta-analyses (Chabris, 1999; Pietschnig et al., 2010) find a tiny-to-nil effect best explained by arousal and mood — anything you enjoy works as well ⚖. There is, so to speak, an equally valid milkshake effect. Verdict. ⚰️ Dead. Mozart is a composer, not a supplement. The enjoyment was already the payment.

⚰️ The 10,000-hour rule

The claim. Ten thousand hours of practice makes anyone an expert at anything. Why it spread. Ericsson, Krampe, and Tesch-Römer (1993) found elite violinists had accumulated on average about ten thousand hours of deliberate practice by twenty; Gladwell (2008) rounded a mean into a law with a number that fits on a poster. What the evidence says. Macnamara, Hambrick, and Oswald's 2014 meta-analysis: deliberate practice explains about 12% of performance variance overall — ~26% in games, ~21% in music, under 1% in professions ⚖. The average hid an enormous spread; hours are an entry fee, not a purchase; and Ericsson himself rejected the "rule" (an average is not a threshold, and deliberate practice — at the edge, with feedback — is not the same as hours). The research the myth mangled is alive and well; it is Chapter 15. Verdict. ⚰️ Dead. The number was a mean. The magic was the feedback loop.

Six graves, one pattern: each myth took a true, narrow result and sold it wide. The autopsy skill is itself transferable — it is 6·35's move in a lab coat: check what the aggregate hides, ask what n was, what task, how long the effect lasted, and who is selling. The next beautiful claim is already on its way.

Further Reading {#further}

Seven doors, honestly labeled.

• Martin Gardner, the "Mathematical Games" corpus (1956–1981). The greatest amplifier in the genre's history; start with The Colossal Book of Mathematics (2001) and budget for the disappearance of several weekends.
• **George Pólya, How to Solve It (1945).** The founding document of heuristics-as-curriculum; Part III of this book is one of its many descendants.
• **Raymond Smullyan, What Is the Name of This Book? (1978).** The knights-and-knaves ladder that walks a patient reader, one tavern riddle at a time, all the way to Gödel.
• **H. E. Dudeney, The Canterbury Puzzles (1907) and Amusements in Mathematics (1917); Sam Loyd, Cyclopedia of Puzzles (1914).** The golden age at the source — Dudeney for elegance, Loyd for showmanship, both read best wearing the provenance goggles Chapter 2 issued.
• Art of Problem Solving — artofproblemsolving.com. Textbooks, forums, and the nursery of most modern olympiad medalists; the working ladder from school mathematics to Chapter 8–11 territory.
• Project Euler — projecteuler.net. Hundreds of problems engineered so brute arithmetic fails and only an insight runs in under a minute; mathematics and programming in a single gym.
• The ARC Prize — arcprize.org. Chollet's grids, built to be easy for humans and hard for machines. Do twenty yourself and feel, first-hand, which side of Part IV's ledger you sit on (standings there dated as of July 2026).

Colophon {#colophon}

This book was made the way it tells you to work.

The fifty-puzzle canon was compiled from primary sources, and the corrected attributions were stated in the text rather than footnoted away: the 15 Puzzle belongs to Noyes Chapman, not Sam Loyd; the zebra belongs to Life International, not Einstein; the Gauss anecdote is told as the legend it is. Sixteen of the canon's solutions were machine-verified by exhaustive enumeration or simulation in July 2026 — breadth-first search for the river crossings, all twenty-four cases of the twelve balls, the harmonic sum of the hundred prisoners, the uniqueness of 6210001000 — because a solution is not "verified" until something that cannot want it to be true has agreed. The worked solutions of Part III were re-derived end-to-end during compilation, and where two independent derivations existed, both were run and made to agree. Every claim about artificial intelligence carries its date — July 2026 — on purpose: those sentences will age, and the date is the difference between a snapshot and a pretense.

The book was composed by the AICE team, July 2026 — humans and machines in the division of labor the book itself recommends: the machines enumerated, checked, and never got tired; the humans chose the puzzles, the order, and what mattered.

One box has stayed open since the first page.

At the gate of Chapter 1, the Sphinx asked: what goes on four legs in the morning, two legs at noon, and three legs in the evening? Sixteen chapters and three appendices, and no answer — a spoiler discipline even we found hard to keep.

The answer is a human being. Crawling on four limbs in the morning of life, striding on two at its noon, leaning on a stick in its evening — one creature, three shapes, a single day standing for a lifetime. Oedipus solved it with the moves you now own: refuse the literal legs, re-bind the words, change the timescale. The Sphinx destroyed herself, as gatekeepers do when the gate is truly passed.

And notice what the answer was. Not a god, not a monster — the animal that changes how it moves and knows it. The oldest recorded riddle is about its own solver: the creature whose whole life is re-representation. Four thousand years later, her newest descendant sits in a benchmark showing colored grids to machines, still sorting minds at the gate. It is her only job. It has always been steady work.

The box is closed; the book is done; the gate, as always, is open.

Go on through.