All Chapters

The Missing Chapter

Gödel's Incompleteness Theorems

Why mathematics will never be able to fully explain itself

An extension of Jordan Ellenberg's "How Not to Be Wrong"

Chapter 53

The Dream of Perfect Knowledge

In 1900, David Hilbert stood before the International Congress of Mathematicians in Paris and essentially said: we can figure out everything. Give me the right axioms, the right rules, and mathematics will become a perfect machine — consistent, complete, and decidable. Every true statement provable. No contradictions. The ultimate flex of human reason.

Hilbert wasn't being naive. He was being ambitious in the way that only a mathematician at the turn of the century could be. The previous decades had been rough for the foundations of math. Cantor had found infinities of different sizes and people were losing their minds about it — some figuratively, Cantor himself literally.1 Russell had discovered that naïve set theory was contradictory (consider the set of all sets that don't contain themselves — does it contain itself?). The whole edifice was wobbling.

Hilbert's program was the repair job. The plan: formalize all of mathematics into a system of axioms and logical rules. Then prove — from outside the system, or from within it — that the system was:

Consistent: You can never prove both a statement and its negation. (No contradictions.)

Complete: Every true mathematical statement can be proven within the system.

Decidable: There exists a mechanical procedure to determine whether any given statement is provable.

This wasn't just intellectual housekeeping. This was a vision of mathematics as a kind of machine: feed in a question, turn the crank, get the answer. Hilbert famously declared, "We must know. We will know."2 It was the motto of his mathematical optimism, and he meant it. He had it inscribed on his tombstone.

Then, in 1931, a quiet 25-year-old Austrian logician named Kurt Gödel published a paper that demolished the whole program in one stroke. The title was modest — "On Formally Undecidable Propositions of Principia Mathematica and Related Systems" — but the content was a bomb. Gödel showed that Hilbert's dream was not just unrealized, but unrealizable. And the way he did it was one of the most beautiful and devious arguments in all of mathematics.

· · ·
Chapter 53

The Trick That Talks About Itself

The key insight — the thing that makes Gödel's proof so fiendishly clever — is self-reference. Gödel figured out how to make a mathematical statement that says something about itself. Not metaphorically. Literally.

Here's the warm-up. You've probably encountered the liar's paradox:

"This sentence is false."

Is it true? If it's true, then what it says is the case, which means it's false. But if it's false, then what it says is not the case, which means it's not false — i.e., it's true. You're caught in an infinite loop. Philosophers have been arguing about this since Epimenides the Cretan said "All Cretans are liars" around 600 BC, and they still haven't entirely sorted it out.3

But Gödel didn't use the liar's paradox directly — that would give you a contradiction, and contradictions aren't useful for proving things. Instead, he pulled an absolutely brilliant substitution. Instead of "This sentence is false," he constructed a sentence that says:

"This statement is not provable."

Think about that for a second. Call this statement G. Now ask: is G provable?

Case 1: G is provable. Then what G says is wrong (it says it's not provable). So we've proved something false. That means our system is inconsistent.

Case 2: G is not provable. Then what G says is true (it correctly asserts its own unprovability). So G is true but unprovable.

Either way, you lose. If your system is consistent (no contradictions), then G is true but can't be proved within the system. That's the First Incompleteness Theorem: any consistent formal system strong enough to do basic arithmetic is necessarily incomplete.

FORMAL SYSTEM Provable Truths G True but unprovable 1+1=2 ∀n: n+0=n ∃ primes > k "I am not provable"
In any consistent formal system, there exist true statements (like G) that lie outside the circle of provable truths.

But wait — you might think, "Fine, just add G as a new axiom!" You can do that. But Gödel's argument applies to the new system too. There's a new Gödel sentence G′ for that system. And on and on, forever. Incompleteness isn't a bug you can patch. It's a feature of any sufficiently powerful formal system.

· · ·
Chapter 53

How Math Talks About Itself

Okay, but I glossed over something crucial. How do you make a mathematical formula say "This statement is not provable"? Mathematical formulas are about numbers, not about themselves. "3 + 5 = 8" doesn't have any opinions about its own provability.

This is where Gödel's real ingenuity comes in: Gödel numbering. The idea is to assign a unique number to every symbol, every formula, and every proof in the entire formal system. It's like a code — every mathematical statement gets a numeric fingerprint.

Gödel Numbering (simplified)
0 1S 2+ 3= 5( 7) 11
Each symbol maps to a prime. Formulas become products of prime powers.

Once every statement has a number, you can write number-theoretic statements that are secretly about other statements. The statement "there is no number that encodes a proof of the statement with Gödel number n" is a perfectly valid arithmetical assertion. It's a statement about numbers. But it's also a statement about provability.

The trick is to find a formula φ(x) meaning "the statement with Gödel number x has no proof," and then construct a sentence whose own Gödel number gets plugged into φ. The result is a sentence that says: "The statement with my Gödel number has no proof" — i.e., "I am not provable."4

It's like writing a computer program that prints its own source code, except instead of printing itself, it asserts something about itself. Self-reference through encoding. The serpent eats its own tail, except the serpent is made of numbers.

🔄 Interactive: The Self-Reference Machine

Build self-referential sentences step by step — from the liar's paradox to Gödel's construction.

Step 1 of 6 — The Liar

Start with ordinary self-reference:

"This sentence has exactly thirty-six letters."

Count them — it really does! Self-reference in English is easy because we have the word "this." But formal math doesn't have "this." So how does Gödel pull it off?

Step 2 of 6 — The Liar's Paradox

Now make it paradoxical:

"This sentence is false."

If true → false. If false → true. An infinite loop with no stable answer. This is the liar's paradox — known since ancient Greece. But Gödel needed something subtler…

Step 3 of 6 — Gödel's Twist

Replace "false" with "unprovable":

"This statement is not provable."

No paradox now! If the system is consistent, this is simply true and unprovable. The swap from "false" to "not provable" is the whole magic.

Step 4 of 6 — But How? Encoding!

Formal math can't say "this." So we encode statements as numbers:

0 = 0→ Gödel number: 2¹ × 3⁵ × 5¹ = 4,050 S(0) = S(0)→ Gödel number: 2² × 3⁷ × 5¹ × 7⁵ × 11¹¹ × 13² × 17⁷ × …

Every formula gets a unique number. Now "the statement with Gödel number 4050" is a way of pointing at "0 = 0" using numbers alone.

Step 5 of 6 — The Provability Predicate

We can write a formula that means "x is not provable":

¬∃p : Proof(p, x)

"There is no number p that encodes a valid proof of the statement with Gödel number x." This is a perfectly normal number theory statement — it just happens to talk about provability.

Step 6 of 6 — The Self-Referential Knockout

Now plug the formula's OWN Gödel number in for x:

¬∃p : Proof(p, ⌜G⌝)

Where ⌜G⌝ is the Gödel number of this very formula. The formula now says: "There is no proof of me." If the system is consistent, this is true but unprovable. That's Gödel's First Incompleteness Theorem. 🎤💧

· · ·
Chapter 53

The Second Blow

Okay, so the First Incompleteness Theorem says there are true statements that your system can't prove. That's bad enough. But the Second Incompleteness Theorem is worse.

Gödel's Second Incompleteness Theorem

No consistent formal system powerful enough for arithmetic can prove its own consistency.

Think about what this means. The one thing you'd most want to know about your axiomatic system — "does it ever contradict itself?" — is the one thing the system is provably unable to tell you. It's like a doctor who can diagnose everyone else's illnesses but is constitutionally incapable of checking whether they themselves are sick.

The proof is elegant. If the system could prove "I am consistent," then it could also prove "The Gödel sentence G is true" (since G is true exactly when the system is consistent). But that would mean it could prove G. But G says it's unprovable. Contradiction — unless the system is inconsistent. So the assumption that the system proves its own consistency leads to inconsistency.5

Hilbert's tombstone reads "Wir müssen wissen. Wir werden wissen." — "We must know. We will know." It was carved in 1943. Gödel's proof had been out for twelve years.

1900 Hilbert's Program 1910 Principia Mathematica 1931 Gödel's Proof 1936 Turing's Halting Problem 1943 Hilbert's tombstone
The timeline of Hilbert's dream and its destruction. Gödel was 25 years old.
· · ·
Chapter 53

Experiencing Incompleteness

All of this might feel abstract. Let me make it concrete with a toy example borrowed from Douglas Hofstadter's Gödel, Escher, Bach.6

The MIU system is a simple formal system. You start with the string MI and try to produce new strings using exactly four rules. The question: can you produce the string MU?

Try it yourself:

🧩 Interactive: The MIU Formal System

Apply the rules to transform strings. Can you reach the target?

Target: MU
MI
1 If string ends in I, add U at the end.   xI → xIU
2 After M, double everything.   Mx → Mxx
3 Replace any III with U.   xIIIy → xUy
4 Remove any UU.   xUUy → xy
MI

If you've been trying to reach MU, you might be getting frustrated. Good. That frustration is the feeling of incompleteness. MU is underivable in the MIU system — and you can prove it using an invariant argument that lives outside the system. Count the number of I's: you start with 1. Rule 2 doubles it. Rule 3 subtracts 3. Rules 1 and 4 don't change the count. To get MU, you need 0 I's. But starting from 1, and only doubling or subtracting 3, you can never reach a multiple of 3 — and you need a multiple of 3 to eliminate all I's via Rule 3. (Proof: 1, 2, 4, 8, 16… modulo 3 gives 1, 2, 1, 2, 1, 2… never 0.)7

That's a miniature Gödel theorem. There's a truth about the system ("MU is not producible") that the system can't express or derive internally — you need to step outside and reason about the system.

· · ·
Chapter 53

The Halting Problem: Gödel Goes Digital

Five years after Gödel's proof, Alan Turing independently discovered what amounts to the same limitation, but in a computational setting.8 Turing's version asks: can you build a computer program that, given any other program and its input, reliably decides whether that program will eventually halt or run forever?

The answer is no, and the proof is structurally identical to Gödel's. Suppose such a Halting Detector H exists. Build a devious program D that uses H to check itself: "Will I halt?" If H says yes, D loops forever. If H says no, D halts immediately. Either way, H gives the wrong answer. Contradiction.

The deep pattern is the same: self-reference creates an unavoidable blind spot. Every sufficiently powerful system — whether it's a formal axiom system or a general-purpose computer — has statements it can express but not resolve.

Gödel "This statement is not provable" If provable → inconsistent If unprovable → true ∴ Incomplete Turing "This program does the opposite" If halts → loops If loops → halts ∴ Undecidable Same structure, different domain
Gödel's incompleteness and Turing's halting problem are structurally the same argument — self-reference creates unavoidable blind spots.
· · ·
Chapter 53

What Incompleteness Does Not Mean

Gödel's theorems are among the most misused results in all of mathematics. Let me be blunt about what they do not say.

They do not mean "math is broken." The vast majority of interesting mathematical statements are perfectly provable. Gödel sentences are very specific, self-referential constructions. Most of the math you know — algebra, calculus, geometry, probability — is completely unaffected.

They do not mean "there are no objective truths." Postmodernists have occasionally waved at Gödel to argue that all knowledge systems are incomplete, therefore everything is relative. This is roughly like arguing that because your car can't fly, transportation is impossible. Gödel's theorems are precise statements about formal systems of a specific type. They don't generalize to poetry or politics.

They do not prove that human minds are superior to computers. Roger Penrose has argued this — that since humans can "see" the truth of Gödel sentences that formal systems can't prove, human consciousness must involve non-computable processes. Most logicians and computer scientists disagree. As Penrose's critics note, humans can't actually verify the truth of a Gödel sentence without first assuming the consistency of the system — exactly the thing Gödel showed can't be proved internally. We don't transcend the limitations; we just outsource them to our intuitions.9

They are not a counsel of despair. If anything, Gödel's theorems tell us something beautiful about mathematical truth: it's bigger than any single formal system can capture. There's always more to discover, always another horizon. Mathematics is not a finite game you can win. It's an infinite one you get to keep playing.

Incompleteness is not a flaw in mathematics. It's a sign that mathematical reality is richer than any formal net we cast into it.

Hilbert wanted a machine that would answer every question. What Gödel showed is that mathematics is not a machine — it's something more wild and open-ended than that. And honestly? That might be the best news mathematicians ever got. Their job security is literally provably guaranteed.

Notes & References

  1. Cantor suffered repeated bouts of depression and was institutionalized multiple times, partly due to the hostile reception of his work by contemporaries like Kronecker. See Joseph Dauben, Georg Cantor: His Mathematics and Philosophy of the Infinite (Harvard University Press, 1979).
  2. Hilbert delivered this address on September 8, 1930 at the Society of German Scientists and Physicians in Königsberg — ironically, the day before Gödel first announced his incompleteness result at the same conference. See Constance Reid, Hilbert (Springer, 1970), pp. 192–196.
  3. For a thorough philosophical treatment of the liar's paradox through history, see JC Beall, Liars and Heaps: New Essays on Paradox (Oxford University Press, 2003).
  4. The precise construction uses the fixed-point lemma (also called the diagonal lemma): for any formula φ(x) with one free variable, there exists a sentence S such that the system proves S ↔ φ(⌜S⌝). See Kurt Gödel, "Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I," Monatshefte für Mathematik und Physik 38 (1931): 173–198.
  5. For a highly accessible treatment of the second theorem, see Peter Smith, An Introduction to Gödel's Theorems, 2nd ed. (Cambridge University Press, 2013), chapters 30–32.
  6. Douglas Hofstadter, Gödel, Escher, Bach: An Eternal Golden Braid (Basic Books, 1979), chapter 1. The MIU system is Hofstadter's toy formal system designed to give readers a feel for formal reasoning.
  7. This is the modular arithmetic argument: starting from 1, the number of I's modulo 3 follows the pattern 1, 2, 1, 2, … under doubling, and subtracting 3 doesn't change the residue mod 3. Since we need 0 mod 3 to eliminate all I's, and we never reach it, MU is underivable. See Hofstadter, GEB, pp. 259–260.
  8. Alan Turing, "On Computable Numbers, with an Application to the Entscheidungsproblem," Proceedings of the London Mathematical Society, Series 2, 42 (1936): 230–265.
  9. For criticisms of Penrose's argument, see Solomon Feferman, "Penrose's Gödelian Argument," Psyche 2 (1995): 21–32; and Martin Davis, "Is Mathematical Insight Algorithmic?" in The Universal Turing Machine: A Half-Century Survey (Springer, 1995).