Truths Math Cannot Prove

Axioms go in, a crank made of pure logic turns, and theorems come out the other side, one after another, forever. Somewhere in that endless output there is one sentence the crank can never print. This lesson is the story of how a 24-year-old found it, why it has to exist in any honest system of mathematics, and what it does and does not mean.

Before you start. You already know what a proof is: a chain of steps from statements you start with to a conclusion, with no step allowed to sneak in something unearned. The only new idea this note asks you to accept up front is that a machine, not just a person, could check every one of those steps.

The idea in 30 seconds

Hilbert wanted one rulebook, checkable by a machine, to settle all of mathematics. Gödel showed that’s impossible, in 1931, at age 24.
Any honest (consistent) rulebook that a machine can check, and that can do ordinary arithmetic, has a true statement it can never prove.
Gödel numbering turns any sentence, and any proof, into a single number, so the rulebook can ask “does this number prove that number” as ordinary arithmetic.
The unprovable sentence says “no number proves me.” It is true, because no such number exists, and unprovable in that same rulebook, because proving it would break consistency.
Patching the gap by adding the sentence as a new axiom never closes it, and the rulebook cannot even prove its own honesty.

Read the story · Skip to the numbers