INSTAGRAM@100xengineers

Liam Price, a 23 year old just solved a math problem on which mathematicians were stuck for the past 60 years using nothing but a $20 ChatGPT subscription. He just casually feeds old math problems into AI to see what happens, something researchers have started calling "vibe maths." So he typed in Erdős Problem , a number theory conjecture open since 1966, into GPT-5.4 Pro. The model spent 80 minutes reasoning and came back with what looked like a complete proof.

The video effectively uses a high-stakes hook to grab attention immediately.
The split-screen format provides visual context for the complex math topic.
The narrative structure follows a classic underdog story arc.

curiosityprofessionalfast cuts#1196

Hook

8/10

Viral

8/10

Views190.0K

Likes17.1K

Analysis

Execution Elements

Visual Style

professional

natural

Audio Design

energy driver

transitions

The Hook

curiosity

High-stakes premise: 23-year-old solves 60-year-old math problem

Setting

home interior

computer monitor, desk chair, coffee mug

Structure Analysis

6 shots

Hook

Introduce the protagonist and the impossible problem he solved.

00:00 - 00:051 SHOT

100:00

medium shoteye-levelstatic

establish the contrast between complex math and the presenter

Camera

static

Action

Split screen: top shows math on chalkboard, bottom shows presenter talking.

Audio

voiceover narration

Problem

Establish the underdog narrative: no degree, no lab, no backing.

00:05 - 00:122 SHOTS

200:05

medium shoteye-levelstatic

visualize the low cost of the solution

Camera

static

Action

Screen recording of ChatGPT subscription page.

Audio

voiceover narration

hard-cut

300:07

medium shoteye-levelstatic

humanize the protagonist

Camera

static

Action

B-roll of Liam Price working at a computer.

Audio

voiceover narration

Solution

Explain the 'vibe maths' method of feeding problems into AI.

00:12 - 00:311 SHOT

400:12

medium shoteye-levelstatic

explain the process

Camera

static

Action

Presenter talking to camera, interspersed with screen recordings of AI math output.

Audio

voiceover narration

Proof

Show the validation from experts and the AI's unique approach.

00:31 - 00:481 SHOT

500:31

medium shoteye-levelstatic

provide expert validation

Camera

static

Action

B-roll of mathematicians at chalkboard, interspersed with presenter.

Audio

voiceover narration

CTA

Encourage following for more AI content.

00:48 - 01:081 SHOT

600:48

medium shoteye-levelstatic

final CTA

Camera

static

Action

Presenter talking to camera, ending with CTA graphic.

Audio

voiceover narration

Script

Full screenplay with dialogue and action

220 words

6 shots

5 beats

Screenplay

6 PAGES

Screenplay

Liam Price, a 23 year old just solved a math problem on which mathematicians were stuck for the past 60 years using nothing but a $20 ChatGPT subscription. He just casually feeds old math problems into AI to see what happens, something researchers have started calling "vibe maths." So he typed in Erdős Problem #1196, a number theory conjecture open since 1966, into GPT-5.4 Pro. The model spent 80 minutes reasoning and came back with what looked like a complete proof.

Page

1 / 6

by 100xengineers

INT./EXT. — HOOK

Introduce the protagonist and the impossible problem he solved.

Split screen: top shows math on chalkboard, bottom shows presenter talking.

[00:00]

HOST

This 23 year old

[00:01]

just solved a math

[00:02]

problem on which

[00:03]

mathematicians were

[00:04]

stuck for the past 60 years

INT./EXT. — PROBLEM

Establish the underdog narrative: no degree, no lab, no backing.

Screen recording of ChatGPT subscription page.

[00:05]

HOST

using nothing

[00:06]

but a $20

B-roll of Liam Price working at a computer.

[00:07]

HOST

ChatGPT subscription.

5 Beats · 6 ShotsSuperdirector.appEN

Screenplay

Page

2 / 6

[00:08]

HOST

Meet Liam Price

[00:09]

A guy with no degree

[00:10]

No research lab

[00:11]

and no institutional backing whatsoever

INT./EXT. — SOLUTION

Explain the 'vibe maths' method of feeding problems into AI.

Presenter talking to camera, interspersed with screen recordings of AI math output.

[00:13]

HOST

He just casually

[00:14]

feeds old math problems into AI

[00:16]

to see what happens,

[00:17]

something researchers started calling vibe maths.

[00:19]

So he

[00:20]

typed in Erdős problem

[00:21]

number 1196,

[00:23]

a number theory

5 Beats · 6 ShotsSuperdirector.appEN

Screenplay

Page

3 / 6

[00:24]

HOST

conjecture open since

[00:25]

1966

[00:26]

into GPT-5.4 Pro.

[00:27]

The model

[00:28]

spent 80 minutes

[00:29]

reasoning and came back

[00:30]

with what looked like a

INT./EXT. — PROOF

Show the validation from experts and the AI's unique approach.

B-roll of mathematicians at chalkboard, interspersed with presenter.

[00:31]

HOST

complete proof.

[00:32]

Here is the wild part.

[00:33]

Every expert who had tried

[00:34]

this problem before

[00:35]

started with the exact

5 Beats · 6 ShotsSuperdirector.appEN

Screenplay

Page

4 / 6

[00:36]

HOST

same approach,

[00:37]

the first move,

[00:38]

90 years straight.

[00:39]

And GPT, having

[00:40]

no idea what the

[00:41]

conventional approach even was.

[00:42]

It found

[00:43]

a completely different technique

[00:44]

using something called

[00:45]

von Mangoldt chains,

[00:46]

which no one

[00:47]

had ever applied

INT./EXT. — CTA

Encourage following for more AI content.

Presenter talking to camera, ending with CTA graphic.

[00:48]

HOST

to this class of problems before.

5 Beats · 6 ShotsSuperdirector.appEN

Screenplay

Page

5 / 6

[00:49]

HOST

Fields Medalist

[00:50]

Terence Tao reviewed it and

[00:51]

called it a

[00:52]

genuine contribution that

[00:53]

leads to the

[00:54]

birth of a new

[00:55]

mathematical theory.

[00:56]

The raw AI output

[00:57]

needed a team

[00:58]

of eight mathematicians

[00:59]

to clean up and formalize,

[01:00]

the key

[01:01]

came from a model

[01:02]

that costs $20 a

[01:03]

month and

5 Beats · 6 ShotsSuperdirector.appEN

Screenplay

Page

6 / 6

[01:04]

HOST

who said

[01:05]

didn't even know what the problem was.

[01:06]

For more such AI content, follow 100x.

— THE END —

5 Beats · 6 ShotsSuperdirector.appEN

Transcript

Monologue

220 words

63 segments

This 23 year old

just solved a math

problem on which

mathematicians were

stuck for the past 60 years

Creator Anatomy

Explore the creator page behind this analysis.

OpenAI

@openai on instagram • 5534782 followers

Turn winning video patterns into your own workflow

Analyze your own account with Superdirector to turn winning hooks into your idea feed and scripts.