J'ai donné 1h à GPT-6 et Opus 5.5 pour recréer GTA 6

J'ai donné 1h à GPT-6 et Opus 5.5 pour recréer GTA 6

I Gave GPT-6 and Opus 5.5 One Hour to Recreate GTA 6

by Ludo Salenne · View the channel on PSI

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 Ludo Salenne 👥 270K 📅 September 28, 2026 ⏱ 27 min 👁 575 📄 science communication 🧭 2026-09-28
Available in: English (current) Français

Keywords

GPT-6Opus 5.5AIGTAgame development

Summary

In this video, Ludo Salenne tests two AI models, GPT-6 Astra and Opus 5.5, by giving them the same prompt to create a GTA-like game and a website with a trailer within one hour. He runs both AIs concurrently with full autonomy. He monitors their progress, noting that both create sub-agents to handle different tasks. Claude Opus 5.5 finishes in about 40 minutes, while GPT-6 takes 50 minutes. The creator compares the final products: the websites, trailers, and gameplay. He finds Opus 5.5’s site and trailer more polished, while GPT-6’s game has slightly better 3D graphics but is less comprehensive. He also analyzes token usage: Opus used only 1% of his $200 plan, while GPT-6 used 13%. He concludes that Opus 5.5 is the better value and more efficient. The video includes live gameplay demonstrations and his honest opinion.

139 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides a practical, hands-on demonstration of AI capabilities in game development. The argumentation is based on direct observation and comparison, with clear criteria: site quality, teaser, gameplay, cost. However, the evaluation is subjective and based on a single trial, so conclusions may not be generalizable. The creator’s writing is engaging and shows the process transparently.

Scientific Rigor, Source Quality, Title Accuracy

The title accurately reflects the content. The video does not cite external scientific sources but relies on its own experiment. The creator discloses the setup (plans, costs) and provides links to the generated websites. However, the methodology is not rigorous enough for a scientific comparison: no controlled variables, single trial, and subjective scoring. The video is more of a demo than a scientific study.

136 words

Title / Content Match

The title accurately describes the content.

Quality & Reliability

7/10

The video presents a hands-on comparison of two AI models with detailed observations, but relies on subjective evaluation and a single trial.

Chapters

Cited Sources

Contribution & Novelties

The video offers a unique comparative playground for cutting-edge AI models, showing their autonomous capabilities in a complex creative task. It demonstrates that AI can generate complete games and websites with minimal human intervention, and highlights cost and performance differences.

Pour aller plus loin :

83 words

Radar Profile

The radar profile shows decent scores in information quantity and quality, but lower in technical depth and reliability, reflecting a practical but non-exhaustive comparison.

Reliability 6/10