Viral Hooks
Trending short videosAI models compete building bridges
R

@rpn

Instagram · Vlog

@rpn Welcome to Episode 02 of Agent Wars. Today we’re challenging 5 top AI models to design, engineer and 3D print a bridge with the goal of holding the most possible weight. For Episode 02, here are the models I tested, with their reasoning level and the harness I used for each: 
🇺🇸 Anthropic: Claude Opus 5.5 (High) 🇨🇳 Kimi: Kimi K3 (High) 🇺🇸 Meta: Muse Spark 1.3 (High)
🇺🇸 OpenAI: GPT 6 Astra (High)
🇺🇸 SpaceXAI: Grok 4.7 (High) Each bridge had specific parameters. It had to span 2 feet, use a maximum 500 grams of filament, and it had to be under 18 hours of printing time. We specified what type of weight and where we would place it on the bridge. 🏆 THE SCORE: Claude Opus 5.5 held an estimated 130 lbs! It took 9 hours and 11 minutes to print 17 parts and used 441 grams of filament. Meta Muse Spark 1.3 held 26.5 lbs. It took 13 hours and 12 minutes to print 49 parts and used 478 grams of filament. OpenAI GPT-6 Astra held an estimated 17.5 lbs. It took 15 hours and 44 minutes to print 29 parts and used 442 grams of filament. SpaceXAI Grok 4.7 did not finish. On assembly it could not stand up by itself. It took 12 hours and 22 minutes to print 29 parts and used 460 grams of filament. Kimi K3 did not finish. It wasn’t engineered correctly so the bridge could not even be assembled. It took 11 hours and 19 minutes to print 35 parts and used 446 grams of filament. 🏅 That means our champion for this episode is Claude Opus 5.5. Not only did it hold the most weight (by FAR) it also took the least time to print, used the least amount of parts and used the least amount of filament.

View original on Instagram

The edit1 shot · 9 text
Shots Text1:19

Performance

Views
3.5M
Likes
148.3K
Comments
584
Engagement
4.2%
Outlier
52.9×
Length · 27 Sept 2026
1:19

Video #43292 · Vlog

AI models compete building bridgesfrom @rpn

[01]

Overview

The creator pits five top AI models (ChatGPT, Claude, Muse Spark, Grok, Kimi) against each other in a challenge to design, engineer, and 3D print the strongest bridge under specific constraints. The video documents the assembly and stress-testing process for each bridge, sequentially eliminating the failures (Kimi, Grok) and ranking the others (OpenAI, Meta) by the weight they can hold. The final and winning bridge, designed by Anthropic's Claude, is tested with over 100 pounds of weights and then with a person standing on it until it dramatically breaks.

[02]

Simple outline

  1. 1State competitive challenge
  2. 2Eliminate low performers
  3. 3Establish performance benchmark
  4. 4Reveal dominant winner
  5. 5Execute surprise final test

[03]

Related videos

[04]

Classification

Format
Concepts
ChallengeQuest RecapProcess
Elements
Text OverlayB-Roll Cutaway360 CameraMontage Editing

[05]

Hook analysis

What stops the scroll in the first 3 seconds

Title hook · 35%

THE FIVE TOP

Verbal hook · 60%

We asked the five top AI models to design, engineer, and three d print a bridge to see which one could hold the most weight.

Visual hook · 5%

A man holds 3D printed bridge components while animated logos of five major AI companies float around him in a workshop setting.

[06]

Hook template

Steal this for your own video

We asked [number of contenders] [competitor type] to [complete the same challenge] to see which one could [win condition].

  • — We asked five AI models to design a bridge to see which one could hold the most weight.
  • — We asked three chefs to make a meal from the same ingredients to see which one could earn the highest taste-test score.
  • — We asked four gardeners to grow tomatoes under the same conditions to see which one could produce the biggest harvest.

[07]

Hook and payoff

The promise the opener makes, and how the video pays it back

Hook

A direct statement of a high-stakes competition between multiple entities, posing a clear question ('which is best?') and promising a definitive, evidence-based answer.

Payoff

The final, official declaration of the champion, validating the entire process and providing a satisfying conclusion to the competition.

[08]

Narrative framework

The structural pattern this video runs on

Framework

The Performance Gauntlet

A systematic, escalating competition where multiple contenders are tested against a single, quantifiable metric, revealing a clear winner through a process of elimination and benchmark-setting, culminating in a surprise final test.

  1. 1

    The Setup

    Establishes the competition's premise, introduces the players, and defines the standardized criteria for judgment to ensure a fair contest and set audience expectations.

  2. 2

    The Culling

    Systematically removes the weakest contenders who fail to meet the basic requirements of the challenge, thereby raising the stakes and emphasizing the difficulty of the task.

  3. 3

    The Benchmark

    Establishes the first quantifiable performance metric with a moderately successful contender, creating a clear baseline that all subsequent participants must surpass.

  4. 4

    The Escalation

    Introduces a superior performer that successfully surpasses the established benchmark, escalating the competition and resetting expectations for what is possible.

  5. 5

    The Champion's Reveal

    Unveils the winning contender by demonstrating performance that dramatically exceeds all previous benchmarks, establishing its clear and undeniable dominance.

  6. 6

    The Victory Lap

    Subjects the winning contender to a surprise, over-the-top final test that goes beyond the original rules, cementing its victory in a memorable and definitive way.

Themes

  • Competition
  • Innovation
  • Artificial Intelligence
  • Engineering Prowess

Techniques

  • Stress Testing
  • 3D Printing
  • AI Design

Trends

  • AI vs AI
  • Practical AI Applications

Post caption

Welcome to Episode 02 of Agent Wars. Today we’re challenging 5 top AI models to design, engineer and 3D print a bridge with the goal of holding the most possible weight. For Episode 02, here are the models I tested, with their reasoning level and the harness I used for each: 
🇺🇸 Anthropic: Claude Opus 5.5 (High) 🇨🇳 Kimi: Kimi K3 (High) 🇺🇸 Meta: Muse Spark 1.3 (High)
🇺🇸 OpenAI: GPT 6 Astra (High)
🇺🇸 SpaceXAI: Grok 4.7 (High) Each bridge had specific parameters. It had to span 2 feet, use a maximum 500 grams of filament, and it had to be under 18 hours of printing time. We specified what type of weight and where we would place it on the bridge. 🏆 THE SCORE: Claude Opus 5.5 held an estimated 130 lbs! It took 9 hours and 11 minutes to print 17 parts and used 441 grams of filament. Meta Muse Spark 1.3 held 26.5 lbs. It took 13 hours and 12 minutes to print 49 parts and used 478 grams of filament. OpenAI GPT-6 Astra held an estimated 17.5 lbs. It took 15 hours and 44 minutes to print 29 parts and used 442 grams of filament. SpaceXAI Grok 4.7 did not finish. On assembly it could not stand up by itself. It took 12 hours and 22 minutes to print 29 parts and used 460 grams of filament. Kimi K3 did not finish. It wasn’t engineered correctly so the bridge could not even be assembled. It took 11 hours and 19 minutes to print 35 parts and used 446 grams of filament. 🏅 That means our champion for this episode is Claude Opus 5.5. Not only did it hold the most weight (by FAR) it also took the least time to print, used the least amount of parts and used the least amount of filament.

Text overlays

  • title · beginning

    Top 5 AI models

  • informational · beginning

    The bridge had to span 2 feet, use a maximum 500 g of filament, and had to land under 18 hours of total printing time.

  • informational · middle

    Model Strength Index

  • branding · end

    CLAUDE OPUS 5.5

  • title · end

    WINNER CLAUDE OPUS 5.5 AGENT WARS EPISODE 2

  • branding · beginning

    Kimi K3 (High)

  • branding · middle

    Grok 4.7 (High)

  • branding · middle

    GPT 6 Astra (High)

  • branding · middle

    Muse Spark 1.3 (High)

Transcript

We asked the five top AI models to design, engineer, and three d print a bridge to see which one could hold the most weight. We put ChatGPT, Claude, Muse Spark, Grok, and China's Kimmy to the test. Each model got the same prompt with specific parameters. The bridge had to span two feet, use a maximum 500 grams of filament, and had to land under eighteen hours of total printing time. So let's see who won. Kimmy was disqualified. It wasn't engineered correctly and good not even be assembled. Grok was a DNF. We're able to assemble it, but it couldn't stand up on its own. Astra fared a lot better. Five pounds, 10 pounds, 15 pounds, but it broke at 20. For what it's worth, I think it was the nicest looking bridge. Up next, Metamuse Spark 1.3, which got to five, ten, 15, and held 20 to beat Astra. Now it got a little wobbly at 25, so we put a Mac Mini on it for a good measure. 26.5 pounds. But it collapsed at 30. So, Claude, what do you got? Well, it ripped past Astra and Muse Spark. It got all the way up to a 100. I mean, we actually ran out of weight. We worse curious. How much more could we add? I think we could stand on it. I mean, I'm one fifty. A hundred rank fifty pounds. Ethan, get up there. And, damn, it actually held Ethan for a split second. So Opus 5.5 is by far and away our champion for episode two. I can't get one fifty.

AI models compete building bridges | Viral Hooks