The Diagonal

Infinity, Incompleteness, and the Limits of Machines

How one simple trick broke mathematics, computation, and The Matrix

 [0] 1  0  1  1  0  1  :
  1 [1] 0  0  1  1  0  :
  0  0 [0] 1  0  0  1  :
  1  1  1 [1] 0  1  0  :
  0  0  1  0 [0] 1  1  :
  1  0  0  1  0 [1] 0  :
  :  :  :  :  :  :  :  :
        \
         \ flip the diagonal
          \ escape the matrix

A Simple Question

Can you make a complete list?

Natural Numbers

1, 2, 3, 4 ...

Yes! Keep counting forever.

Pick any number. We'll reach it.

Even Numbers

2, 4, 6, 8 ...

Yes! Every even is covered.

Pick any even. We'll reach it.

Infinite Sequences

1010101...
1111000...
0010110...
1100101...
???????...

Each one goes on forever...

Can you list them all?

Georg Cantor answered this in 1891. His trick is beautiful.

Cantor's Setup

Proof by contradiction: Suppose someone hands you a list and claims it contains every infinite binary sequence.

(This is just an example list - we'll show that any such claimed list must have gaps.)

1.1010100...
2.1111001...
3.0010110...
4.0111001...
5.1000010...
6.0101011...
7.1111010...← the diagonal!
...

Cantor says: "I can always find one you missed."

The Diagonal Move

Diagonal: _ _ _ _ _ _ _ ... (this IS row 7!)
Flip: _ _ _ _ _ _ _ ... differs at position 7!
Escapee: _______...

Why can't the escapee be on the list?

  • Row 1 has 1 at position 1, escapee has 0 → differs!
  • Row 2 has 1 at position 2, escapee has 0 → differs!
  • Row 7 (the diagonal!) has 0 at position 7, escapee has 1 → differs!
  • Row n has X at position n, escapee has not-X → differs!

The escapee differs from every row at exactly the position where it would need to match!

What This Means

Some infinities are bigger than others.

Countable Infinity ℵ₀

Natural numbers: 1, 2, 3, ...

Integers: ..., -2, -1, 0, 1, 2, ...

Fractions: 1/2, 2/3, 7/11, ...

Infinitely many, but each one is finite

 1
 2
 :
 9
 1 0
 1 1
 :
 9 9
 1 0 0
 :

Jagged - diagonal runs off the edge

Uncountable Infinity

Infinite binary sequences: 101010...

Real numbers: 0.14159...

Points on a line

Infinitely many, and each one is infinite

 0 1 0 1 1 :
 1 1 0 0 1 :
 0 0 0 1 0 :
 1 1 1 1 0 :
 0 0 1 0 0 :
 : : : : : :

Square - diagonal escapes forever

The diagonal is a machine that takes any list of infinite sequences and produces one that's missing.

Enter Gödel (1931)

"What if I use the diagonal trick on mathematics itself?"

The Key Insight

Mathematical statements can be encoded as numbers.

1. Give each symbol a code:
"1"→2 "2"→3 "+"→4 "="→5
2. Encode "1 + 1 = 2" using primes:
Symbol1+1=2
Code24253
Prime235711
=3⁴7⁵11³
3. Multiply → one unique number:
"1 + 1 = 2" = 2² × 3⁴ × 5² × 7⁵ × 11³

Every statement is a number. So math can talk about its own statements.

Math talks about numbers: "7 is prime"
Statements ARE numbers: "7 is prime" = 4,827,619
Math talks about statements: "Statement #4827619 is provable"
A statement can talk about itself...

The Self-Referential Escape

Gödel constructs statement G with Gödel number g:

G

"Statement #g is not provable"

But G is statement #g... so G says: "I am not provable"

If G is provable...

Then G is false (it claims to be unprovable)

But we proved something false!

System proves lies. Broken.

If G is not provable...

Then G is true (it correctly claims its unprovability)

A true statement escapes!

System is incomplete. Always.

Cantor: "List all sequences" → Diagonal escapes the list

Gödel: "Prove all truths" → G escapes the proofs

Turing's Turn (1936)

Can a computer determine if any program will eventually stop?

for i in 1 to 10: print(i) # STOPS

Halts ✓

while True: continue # FOREVER

Never halts ✗

if HALT_DETECTOR(myself): loop_forever() else: stop()

???

The same diagonal trick: Make a program that does the opposite of what the detector predicts.

If the detector says "halts" → loop forever

If the detector says "loops" → halt immediately

No halt detector can exist. Computation has limits.

The Pattern

Self-reference + Enumeration = Transcendence

Enumeration

Try to list/contain everything

+

Self-reference

System can describe itself

=

Transcendence

Something always escapes

Thinker System The Diagonal What Escapes
Cantor List of sequences Flip the diagonal A sequence not on the list
Gödel Formal mathematics G: "I am unprovable" Truths beyond proof
Turing Computation Self-contradicting program Undecidable problems
??? The Matrix Neo (the anomaly) Unpredicted choice

It's the same ghost appearing in different machines.

The Matrix as Formal System

The Matrix is:

  • A simulation (formal system)
  • Bounded by rules, by code
  • Tries to be complete
  • Even includes the "escape" (Zion, the One cycle)

Neo is:

  • The diagonal (the mechanism)
  • The anomaly, generated BY the system
  • Can see and manipulate the code
  • NOT what escapes - what ENABLES escape

"The anomaly is systemic, creating fluctuations in even the most simplistic equations."

— The Architect

Neo is like G - the self-referential mechanism. What escapes is unpredicted choice.

Why The Machines Can't Win

They try to patch it

New version of the Matrix

Creates new anomalies

They try to reset it

Destroy Zion, reload the One

The cycle repeats (6 times!)

They try to control it

Make the "escape" part of the system

Neo makes an unpredicted choice

Zion isn't the escape—Zion is part of the controlled cycle.

Neo isn't what escapes—Neo is the mechanism that enables escape.

What escapes is genuine choice—consciousness that transcends prediction.

What About AI?

What's a formal system?

A system with symbols, rules, and outputs - all precisely defined.

Cantor's list: sequences + enumeration rules Gödel's math: axioms + proof rules Turing's computer: states + transition rules

Is an AI like ChatGPT a formal system?

Yes. An AI has:

  • Symbols: tokens, numbers, embeddings
  • Rules: matrix math, activations, layers
  • Outputs: input → computation → response
       ┌─────────┐   0.7   ┌─────────┐
   ──▶ │ THINK   │ ──────▶ │ ANSWER  │
       └────┬────┘         └─────────┘
            │ 0.3
            ▼
       ┌─────────┐   0.8   ┌─────────┐
       │ RETRY   │ ──────▶ │ ANSWER  │
       └─────────┘         └─────────┘

Same input, different paths. It's a probability machine (NFA).

So Gödel and Turing apply:

  • It cannot fully model itself
  • It cannot predict all its own outputs
  • There will always be questions it cannot answer

But here's the thing...

The Dance

Humans are ALSO formal systems.

We ALSO have Gödelian limits.

When a human talks to an AI:

  • Two incomplete systems
  • Each seeing things the other misses
  • Each being the other's diagonal sometimes
  • Together, reaching further than alone
    Human          AI
      \           /
       \         /
        \       /
         \     /
          \   /
           \ /
            X  ← something new emerges
           / \
          /   \
         /     \

The limit isn't a wall. It's where the interesting stuff happens.

Let's See It In Action

Here's a real AI system - a state machine that thinks:

What you'll see:

  • States - what the AI is doing (understanding, designing, implementing)
  • Transitions - decisions that move between states
  • Thinking - the AI's internal reasoning (yes, really)
  • Output - what it produces

Notice: it loops, retries, fails, recovers. It's not magic. It's a system with limits, feeling its way forward.

Your Turn

(click each card to see examples)

Challenge 1: Be Cantor

I'll give you a list of sequences. Find the diagonal escapee.

Example:

1: 1 0 1 0 ...
2: 0 0 1 1 ...
3: 1 1 1 0 ...
4: 0 1 0 0 ...

Diagonal: 1,0,1,0

Flip it: 0,1,0,1

Challenge 2: Be Gödel

Write a sentence that talks about itself. Can you make a paradox?

Examples:

"This sentence has five words."

"This sentence is false." (paradox!)

"I am lying right now."

Can you write your own?

Challenge 3: Be Neo

What are the rules of your school? What "anomalies" exist?

Think about:

Rules that contradict each other?

Things everyone does but aren't "allowed"?

People who don't fit the categories?

Where do the rules break down?

Challenge 4: Be a Scientist

Ask the AI visualizer a question. Watch it think. Where does it struggle?

Try asking:

"What can't you do?"

"Are you conscious?"

"Describe yourself describing yourself"

Watch for loops and hesitation!

The Takeaway

1

Every system has limits. This isn't a bug—it's a mathematical certainty.

2

The diagonal is a tool. It finds what escapes. Learn to use it.

3

AI is powerful but bounded. So are we. The magic is in the collaboration.

4

Reality isn't just math. It's math plus whatever escapes the math.

"There is no spoon."

Translation: The limits of the system are not the limits of what's possible.

Questions?

    _______________
   |  ___________  |
   | |           | |
   | |   ?   ?   | |
   | |     ?     | |
   | |   ?   ?   | |
   | |___________| |
   |_______________|
         |||
    the question machine
    (also has limits)

Let's discuss. The best questions are the ones that break the system.

Thinker System The Diagonal What Escapes
??? The Matrix Neo (the anomaly) Unpredicted choice

They weren't special because of who they were.

They were special because they saw.

Cantor saw through infinite sets.

Gödel saw through formal logic.

Turing saw through computation.

Neo saw through the simulation.

The ??? is an empty seat.