ChatGPT 5.5 Codex : c'est vraiment la révolution que tout le monde attend?

Value of the Information & Strength of the Argument

The video provides valuable practical insights into the limitations of ChatGPT 5.5’s agentic features, particularly the need for explicit memory configuration and the risks of uncontrolled agent generation. The argumentation is persuasive, using a live demonstration to show how the model behaves without proper setup. However, the value is diminished by the heavy promotion of the creator’s paid courses, which are repeatedly inserted into the content. The technical explanations are simplified, which may be suitable for beginners but lack depth for advanced users. The argument that simple prompts are insufficient is well-supported, but the video does not provide a comprehensive solution, instead pointing to paid resources.

Scientific Rigor, Source Quality, Title Accuracy

The video does not cite any external sources or official documentation, relying solely on the creator’s own testing and observations. The title suggests a review of ChatGPT 5.5’s revolutionary potential, but the content is more of a tutorial with a critical perspective. The lack of citations reduces the scientific rigor, and the promotional nature of the video undermines its credibility. The creator’s claims about model behavior are not backed by independent benchmarks or references. The title is somewhat clickbait, as the video does not fully address the ‘revolution’ question but rather focuses on practical usage.

The title is somewhat misleading: the video focuses on a tutorial for creating AI agents with ChatGPT 5.5 Codex, rather than a comprehensive review of whether it is a 'revolution'. The content does address the question indirectly but the main thrust is practical guidance.

6/10

The video provides a practical demonstration of ChatGPT 5.5's agentic features, highlighting limitations and the need for explicit configuration. However, the presentation is heavily promotional, with repeated plugs for the creator's paid courses, and lacks citations to official documentation or independent benchmarks. The technical explanations are simplified and sometimes vague, reducing the overall reliability.

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

• 00:00Introduction: promise of ChatGPT 5.5 for creating AI agents, overview of the tutorial.
• 02:00Explanation of how AI agents work: context window, memory blocks, and coherence issues.
• 05:00Discovery of the orchestrator system and memory files (memory.md, rollout summaries).
• 08:00Demonstration of a simple prompt without memory configuration, showing lack of control.
• 12:00Analysis of the model's behavior: it generates agents autonomously but without user control.
• 16:00Warning about asking AI to do impossible tasks (e.g., accessing closed platforms).
• 20:00Example of a marketing research agent prompt, highlighting unrealistic expectations.
• 25:00Discussion on the need for explicit instructions and memory management for reliable agents.
• 30:00Conclusion: emphasis on critical thinking and avoiding simplistic AI solutions.

• Parlons IA formations — Mentioned as the creator's training platform for AI and business.
• Dailymotion channel — Alternative video platform for the creator's content.
• Medium blog — Blog with additional AI-related articles.
• Podcast — Podcast link for further listening.
• SEO Agent IA tool — Promoted tool for AI agents, with discount code.

• OpenAI documentation on agents — Official documentation that describes agentic patterns and memory management, aligning with the video's technical points.

• OpenAI marketing materials — OpenAI's marketing suggests that creating agents is as simple as a few sentences, which the video argues is misleading based on its testing.