J'ai passé 100h sur GPT-4.5 - Voici comment l'utiliser !

J'ai passé 100h sur GPT-4.5 - Voici comment l'utiliser !

by Ludo Salenne

Humanities, Social Sciences & Thought AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 Ludo Salenne 👥 267K 📅 March 6, 2025 ⏱ 39 min 👁 42K 📄 tutorial 🧭 2026-08-21
Available in: English (current) Français

Keywords

GPT-4.5prompt engineeringchain of draftpersona promptingtutorial

Summary

In this tutorial, Ludo Salenne shares his experience and best practices for using GPT-4.5 after spending 100 hours testing it. He addresses common criticisms of the model, arguing that it is powerful when used correctly. The video covers several key techniques: structuring prompts with important instructions at the beginning and using separators, attributing roles (persona prompting) to guide the model, and using the SMART method for prompt design. He also introduces the concept of ‘chain of draft’ as a more efficient alternative to chain-of-thought prompting, and discusses the importance of breaking complex tasks into smaller steps. The video includes demonstrations of prompt examples and comparisons of outputs with different roles. It also contains a promotional segment for his training course and a clone AI message.

125 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides practical, actionable advice for improving interactions with GPT-4.5. The demonstrations of persona prompting and the comparison of outputs are valuable and illustrate the points effectively. The argumentation is based on personal experience and references to official documentation, but it lacks rigorous scientific validation. The creator’s enthusiasm is evident, but some claims, such as the model’s ‘unsupervised learning’ leading to better understanding, are oversimplified and not fully substantiated.

Scientific Rigor, Source Quality, Title Accuracy

The video cites official OpenAI resources (system card, arXiv paper) and the creator’s own website. The sources are relevant and add credibility. However, the video also includes promotional content for the creator’s training course, which is clearly separated but still present. The title accurately reflects the content, and the video is well-structured with clear chapters. The creator’s personal experience is a valid source of information, but it is not a substitute for peer-reviewed research.

159 words

Title / Content Match

The title accurately reflects the content: the creator shares insights and tips from extensive testing of GPT-4.5.

Quality & Reliability

6/10

The video is a practical tutorial based on personal experience and official OpenAI documentation. It references the GPT-4.5 system card and an arXiv paper, but the claims about model behavior are anecdotal and not systematically verified. The promotional segments and subjective assessments lower the overall reliability.

Chapters

Cited Sources

  • GPT-4.5 System Card — Official OpenAI documentation on GPT-4.5, referenced for technical details.
  • GPT-4.5 System Card PDF — PDF version of the system card, cited for in-depth analysis.
  • Chain of Draft paper — arXiv paper introducing the 'Chain of Draft' prompting method, discussed in the video.
  • Trustpilot reviews — Customer reviews of the creator's training, linked in the description.
  • Formation ChatGPT — Creator's paid training course, promoted in the video.
  • Ressources IA gratuites — Free prompts and GPTs offered by the creator.

Concurring Sources

  • OpenAI GPT-4.5 System Card — Official documentation that supports the video's claims about model capabilities and training.
  • Chain of Draft paper — The paper provides a scientific basis for the 'chain of draft' technique recommended in the video.

Dissenting Sources

  • Community criticism of GPT-4.5 — The video addresses widespread criticism of GPT-4.5's performance, which contrasts with the creator's positive assessment. This criticism is not formally documented in a specific source but is referenced in the video's introduction.

External References

Contribution & Novelties

The video offers a practical, experience-based guide to using GPT-4.5, highlighting specific techniques such as prompt structuring, persona prompting, and the chain of draft method. It provides concrete examples and comparisons that are useful for users seeking to improve their interactions with the model. The emphasis on adapting to GPT-4.5’s unique characteristics (e.g., importance of initial instructions) is a valuable contribution to the community.

Pour aller plus loin :

  • Chain of Draft paper — The academic paper behind the ‘chain of draft’ method, directly relevant to the video’s advice.
  • Prompt engineering guide (OpenAI) — Official OpenAI documentation on prompt engineering, providing broader context.
  • Persona prompting — Wikipedia article on prompt engineering, including persona pattern, relevant to the video’s second tip.
  • SMART criteria — Wikipedia article on SMART criteria, which the video adapts to prompt design.

135 words

Radar Profile

The radar profile shows high scores in information quantity and technical level, reflecting the video's detailed and practical content. However, the quality and reliability scores are moderate, indicating that while the information is useful, it is based on personal experience and promotional elements rather than rigorous scientific analysis.

Reliability 6/10

💬 Très positif. Sur les 30 commentaires analysés, les spectateurs expriment une grande satisfaction, remercient le créateur pour la qualité pédagogique et demandent une vidéo complémentaire sur le choix du modèle selon les cas d'usage.