跳到正文
Full Value Dan· 来源· @fullvaluedan · X·· 6 小时前AI 评分58
正文 · 原文

Can frontier AI build a playable video game from a single TikTok clip?

In this benchmark, we tested the top frontier AI models:
- @claude Opus 5.5
- @ChatGPT Sol 6.1
- @Muse Wonder Full 1.0
- @geminiapp Argon 4.0

The challenge: recreate the Rick and Morty concept where fighters spend the whole game commuting through the city just to reach their street fight.

We ran two distinct rounds:

The "Bad Prompt" Test: A short, lazy prompt with only a reference video link to see which model can infer context and handle poor instructions.

The "Good Prompt" Test: An advanced, structured prompt using Three.js specifications, game rules, and strict mechanics.

Watch to see why prompt engineering is still 80% of the build process, how Gemini Argon performed across three separate platforms (Anti-Gravity, Google AI Studio, and the Gemini App), and which model built the ultimate 3D playable game.

TIMESTAMPS:
00:00 – Can Frontier AI Build the Viral Rick & Morty Game?

00:47 – The "Bad Prompt" Test: 1 Vague Prompt Across 4 Frontier Models

01:11 – Gemini Argon Test: Anti-Gravity vs AI Studio vs Gemini App

05:42 – ChatGPT 6.1 Soul Test: Multimodal Video Scraping & Failure Catch

09:24 – Claude Opus 5.5 Test: Building "Street Grudge"

11:10 – Meta Muse Wonderful Test: Hermes Agent & 15-Minute Game Build

12:47 – Bad Prompt vs Good Prompt: Why Prompting is 80% of the Work

14:55 – Claude Opus 5.5 Prompting Breakdown: 3D Three.js Spec

15:56 – Claude Opus 5.5 3D Game Build Result (Sal Volcano vs Big Lasagna)

18:13 – ChatGPT 6 Prompting Tips & Playable 3D City Build (The Winner)

19:56 – Gemini Argon 3D Game Test: Collision Glitches & UI Slop

22:14 – Meta Muse Wonderful Test: Hermes Agent 1-Hour Build

24:26 – ChatGPT 6 vs Claude vs Gemini Argon: Final AI Coding Rankings

🔗 Twitter: https://twitter.com/fullvaluedan

[Scam Warning] All comments/replies that lead you to Whatsapp, Telegram, Instagram is going to be a scam. Please be careful and never trust anyone with your personal information, including me (because it might be a scammer).

来源:Full Value Dan · x.com