J’ai passé 300h sur GPT-5 – Voici comment bien l’utiliser !

J’ai passé 300h sur GPT-5 – Voici comment bien l’utiliser !

by Ludo Salenne

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 Ludo Salenne 👥 266K 📅 October 10, 2025 ⏱ 31 min 👁 34K 📄 tutorial 🧭 2026-08-03
Available in: English (current) Français

Keywords

GPT-5ChatGPTprompt engineeringRCT methodCARTEL method

Summary

In this tutorial, Ludo Salenne shares his experience of using GPT-5 for over 300 hours and provides five tips to get the most out of it. He begins by noting that GPT-5 can be ’lazy’ and explains that the default ‘Auto’ mode uses a router that may direct requests to less capable models, leading to subpar responses. His first tip is to explicitly ask GPT-5 to ’think carefully’ before answering, which triggers a longer chain-of-thought and often results in higher-quality responses. The second tip is to use the RCT (Role, Context, Task) prompt structure to provide clear context and improve output. He then introduces his own advanced prompt framework called ‘CARTEL’ (though he does not fully explain the acronym in the transcript), which he claims is highly effective. He also demonstrates an example prompt and discusses the importance of using the right model for the task. Additionally, he covers how to use GPT-5 without a subscription, how to automate agents, and highlights four essential features. Throughout the video, he promotes his paid course on automating ChatGPT, which is a significant portion of the content. The tutorial is practical and aimed at helping viewers improve their use of GPT-5, but it is heavily intertwined with promotional material.

206 words

Critical Evaluation

The video provides a practical, experience-based tutorial on using GPT-5, with a focus on prompt engineering techniques to overcome perceived laziness and improve response quality. The author’s main argument is that GPT-5’s default ‘Auto’ mode may route requests to less capable models, but by explicitly requesting deeper thinking and using structured prompts, users can unlock its full potential. This is a plausible and useful insight, as it aligns with known behaviors of large language models where prompt phrasing can significantly influence output. However, the video lacks scientific rigor: the claims are based on anecdotal evidence and personal testing, with no systematic evaluation or comparison against other models. The author does not provide any quantitative data or reproducible experiments, which limits the reliability of the conclusions. The sources cited are mostly links to the author’s own products and other videos on his channel, which are not independent or authoritative. The promotional segment for his paid course is extensive and may bias the content, though it does not invalidate the practical tips. The video is well-structured with clear chapters and visual aids, making it accessible to a general audience. The adéquation between title and content is good, as the author indeed shares insights from 300 hours of use. Overall, the video is useful for practical users but should be viewed with caution due to its promotional nature and lack of scientific validation.

230 words

Title / Content Match

The title accurately reflects the content: the author shares insights and tips from 300 hours of using GPT-5, as promised.

Quality & Reliability

6/10

The video is a practical tutorial based on the author's extensive personal experience with GPT-5. It provides actionable tips and demonstrates techniques, but relies heavily on anecdotal evidence and promotional content for a paid course. The scientific rigor is limited, and the claims about GPT-5's performance are not backed by systematic testing or external validation.

Chapters

Cited Sources

  • Trustpilot Reviews for SLN Web — Cited as evidence of the quality of the author's paid course.
  • Formation ChatGPT (paid course) — Promoted in the video as a comprehensive training on ChatGPT, including GPT-5.
  • Ressources IA gratuites — Mentioned as a source for free AI resources, prompts, and GPTs.
  • OpenAI Chat Platform — Referenced as the platform where GPT-5 is used.

Concurring Sources

  • OpenAI GPT-5 documentation — Official documentation that may confirm the existence of multiple GPT-5 variants and router behavior.

Dissenting Sources

  • Community discussions on GPT-5 performance

External References

Contribution & Novelties

The video offers a practical, experience-based guide to using GPT-5, with a novel emphasis on the router mechanism and the importance of prompting for deeper reasoning. The author introduces his own ‘CARTEL’ method for prompt engineering, which, while not fully detailed in the transcript, appears to be a structured approach to crafting effective prompts. This adds to the existing body of prompt engineering techniques.

Pour aller plus loin :

  • Prompt engineering — Wikipedia article on prompt engineering, providing background on techniques like role prompting and chain-of-thought.
  • Chain-of-thought prompting — Research paper introducing chain-of-thought prompting, which is relevant to the author’s advice on asking GPT-5 to think carefully.
  • OpenAI GPT-5 documentation — Official documentation for OpenAI models, including GPT-5, which can help verify capabilities and usage.

125 words

Radar Profile

The radar profile shows moderate scores across all dimensions, with slightly higher scores in information quantity and technical level, but lower in reliability. This indicates a tutorial that is informative and practical but lacks scientific rigor and independent validation.

Reliability 5/10