Les scientifiques révèlent la meilleure technique de prompt IA en 2026

Value of the Information & Strength of the Argument

The video provides a critical perspective on common prompting advice, arguing that many techniques are outdated. The argumentation is based on a claimed study of 1500 articles, but the specific study is not clearly cited, reducing credibility. The presenter’s personal anecdotes and promotion of his own courses may bias the content. The discussion of model evolution from completion to agentic models is interesting but lacks depth and precise technical details.

Scientific Rigor, Source Quality, Title Accuracy

The video references a study called ‘Harvx’ but does not provide a clear citation or link. The description includes links to the creator’s own platforms and affiliate links, but no direct scientific sources. The title is somewhat misleading as it promises a ‘best technique’ but the content is more of a critique of old methods. The video does not provide a balanced view, and the presenter’s claims are not independently verified.

The title promises a 'best AI prompt technique' but the video primarily critiques old prompting methods and discusses model evolution, with limited concrete new techniques.

4/10

The video presents a mix of personal claims and references to studies, but lacks precise citations and verifiable data. The presenter's expertise is asserted but not demonstrated with credentials. The content is largely opinion-based, with some references to scientific studies that are not clearly identified.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction: presenter claims to have studied 1500 articles on prompt engineering.
• 01:00Discussion of a 2023 study on job classification with different prompting methods.
• 03:00Explanation of the 'perfect prompt' from 2023, including role, repetition, and encouragement.
• 05:00Critique of common prompting advice on social media, calling it outdated.
• 07:00Introduction of 'GDP Val' models trained on professional tasks, potentially impacting jobs.
• 09:00Personal anecdote about using AI for medical assistance.
• 11:00Analysis of ChatGPT 5.2's critique of the 2023 'perfect prompt'.
• 13:00Recommendations for modern prompting: neutral, concise, with clear variables and decision rules.
• 15:00Discussion of the role of role prompting and its limitations.
• 17:00Conclusion: emphasis on adapting to new AI architectures and avoiding outdated methods.

• Parlons IA - Formations — Promotion of the creator's own AI training courses.
• Parlons IA - Dailymotion — Alternative video platform for the channel.
• Parlons IA - Medium Blog — Blog with additional content.
• Parlons IA - Podcast — Podcast link.
• SEO Agent IA — Affiliate link for an AI tool.

• Prompt engineering - Wikipedia — General reference on prompt engineering.

• No specific discordant sources found — The video's claims are not directly contradicted by any specific source, but the lack of clear citations makes it difficult to verify.