J'ai refait Pokémon avec Claude Opus 4.8 en un seul prompt

Value of the Information & Strength of the Argument

The video provides valuable hands-on insights into the practical capabilities of Claude Opus 4.8, especially its new UltraCode mode. The creator demonstrates real-world performance on diverse tasks, which is more informative than benchmark numbers alone. The argumentation is based on direct observation and comparison, but it is subjective and lacks controlled conditions. The creator acknowledges limitations, such as the high token consumption and incomplete results, which adds credibility. However, the evaluation criteria are not clearly defined, and the conclusions are based on personal preference rather than objective metrics.

Scientific Rigor, Source Quality, Title Accuracy

The video cites official claims from Anthropic about reduced hallucinations and improved efficiency, but does not provide direct links to these sources. The creator mentions benchmarks like SWE-bench and Terminal Bench, but does not give specific URLs. The description includes a link to a form for prompts, which is not a scientific source. The title accurately reflects the content, focusing on the Pokémon recreation, which is the highlight. The video is more of a practical demonstration than a rigorous scientific analysis, so the scientific rigor is moderate.

The title accurately reflects the main content: the creator uses Claude Opus 4.8 in a single prompt to recreate a Pokémon game, with additional comparative tests.

6/10

The video is a hands-on test of Claude Opus 4.8, providing practical demonstrations and comparisons with other models. However, the methodology is informal, lacks rigorous controls, and relies on subjective impressions. The creator mentions benchmarks and official claims but does not provide detailed sources or verification.

• 00:00Intro : Claude Opus 4.8 & Mythos
• 05:03Test 1 : Créer un jeu 2D
• 07:03Test 2 : Créer un Windows bis
• 09:03Test 3 : Créer un site web esthétique
• 10:35Test 4 : Simulateur d'avion 3D
• 12:20Test 5 : Reproduire Pokémon sur Visual Studio Code
• 15:27Conclusion et outro

• Prompts pack (free download) — The creator offers the prompts used in the tests for free via this form.

• Anthropic's official claims about Opus 4.8 — The creator mentions Anthropic's claims about reduced hallucinations and improved efficiency, which are consistent with the video's observations.

• Benchmark comparisons — The video shows that Opus 4.8 does not always outperform GPT-5.5 or Qwen 3.7 Max, which contrasts with some benchmark rankings.