1 MILLIARD de tokens et 34 HEURES pour créer ce jeu avec FABLE 5.2 !

1 MILLIARD de tokens et 34 HEURES pour créer ce jeu avec FABLE 5.2 !

1 BILLION tokens and 34 HOURS to create this game with FABLE 5.2!

Title history
  • 1 MILLIARD de tokens et 34 HEURES pour créer ce jeu avec FABLE 5.2 ! — observed on 20261110-1652 UTC (YouTube)
🎙 DevArt 👥 19K 📅 November 10, 2026 ⏱ 14 min 👁 192 📄 documentary 🧭 2026-11-10
Available in: English (current) Français

Keywords

tokensgame developmentAI agentGodotbudget

Summary

In this video, the creator presents a game developed over 34 hours using AI tools, specifically Claude Opus 5.5 Max and Fable 5.2, with a total of 928 million tokens (nearly one billion) and an estimated cost of around €500. The game is a driving simulation set in a 1970s-style city, featuring detailed vehicles, buildings, and dynamic environments. The creator demonstrates gameplay, including driving, refueling, and exploring the town, while highlighting the graphical quality and physics. He explains that he used AI agents with Blender and Godot Engine to generate assets and code autonomously, reviewing and iterating as needed. The video breaks down the cost distribution: 20% for buildings and facades, 15% for the gas station and animations, and 10% each for vehicle modeling, traffic, and other elements. Challenges included achieving real-time visual fidelity and handling bugs like faulty shaders and collision issues. The creator compares this project to an earlier taxi game, noting that the taxi game was larger and cheaper despite impressive rain effects. He concludes by sharing his intention to continue pushing the limits of AI-assisted game development and offers his prompt template for free through a community link.

192 words

Critical Evaluation

Value of the Information & Strength of the Argument

The value of the information lies in providing concrete, quantitative data about the resources required for AI-assisted game development, including token counts, costs, and time. The creator argues that AI can generate a playable game with minimal human intervention, although the quality is not yet production-ready. His argumentation is based on personal experience rather than systematic testing, but he transparently discusses limitations and bugs encountered. He compares with a previous project to contextualize costs, though the comparison is anecdotal. Overall, the video offers a practical glimpse into current AI capabilities for game prototyping, but the evidence is not rigorous enough to draw general conclusions.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is limited; the video is a demonstration rather than a controlled experiment. Sources are minimal: the creator references his own experience and provides links to his community for resources. No external studies or official documentation are cited. The title is accurate and aligned with the content, as the video explicitly mentions the token count and duration. The description includes links to related videos and a free community, but these are promotional rather than substantive sources. The lack of peer-reviewed references or reproducibility details reduces the overall reliability.

209 words

Title / Content Match

The title accurately reflects the main claim: the creator used nearly one billion tokens and 34 hours to build a game with Fable 5.2. The video indeed showcases that effort.

Quality & Reliability

5/10

The video presents a personal demonstration of AI-assisted game development, with specific metrics (token usage, cost, duration) but relies on anecdotal evidence and lacks external verification or comparative analysis.

Key Moments

Cited Sources

Contribution & Novelties

The video contributes a practical case study of using AI agents to create a playable open-world game from scratch, with detailed cost and token metrics. It offers a breakdown of where the AI effort is spent (assets, physics, environment) and highlights common pitfalls. The creator also shares his prompt methodology, which could be reused by others.

Pour aller plus loin :

  • Godot Engine — Open-source game engine used in the project; useful for understanding the underlying technology.
  • Claude by Anthropic — The AI model family (Opus 5.5) used for code generation and asset creation.
  • Vibe coding — A term describing the practice of using AI to generate code interactively, relevant to the creator’s workflow.

114 words

Radar Profile

The radar shows a moderate profile with higher scores in information quantity and technical level, while reliability and quality are lower, indicating that the content is rich in anecdotal data but lacks rigorous verification.

Reliability 4/10