Réponses IA : comment passer de ça... à ÇA !

Réponses IA : comment passer de ça... à ÇA !

by Ludo Salenne

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 Ludo Salenne 👥 266K 📅 May 27, 2026 ⏱ 19 min 👁 15K 📄 tutorial 🧭 2026-08-02
Available in: English (current) Français

Keywords

HTMLAI responsesformattingpromptingproductivity

Summary

The video addresses the problem of AI responses being presented in plain text or basic markdown, which is tiring to read and less engaging. The creator introduces a technique, popularized by Andrej Karpathy, of asking AI models to format responses as standalone HTML documents. This method improves readability, structure, and professionalism. The video explains the technical benefits: better response structuring, access to higher-quality training data, and token efficiency. It demonstrates the technique on various AI tools (Claude, ChatGPT, Gemini) and provides practical use cases: creating synthesis sheets, comparison tables, training guides, and professional reports. The creator emphasizes that this simple addition to prompts can transform the user experience and make AI outputs more actionable. The video includes promotional segments for the creator’s training and community.

125 words

Critical Evaluation

The video presents a practical and timely technique for improving AI output readability, which is valuable for both developers and non-developers. The demonstration is clear and the before/after comparisons are effective in illustrating the benefits. The argumentation is solid, relying on the authority of Andrej Karpathy and the inherent advantages of HTML formatting. However, the scientific rigor is limited: the claims about better corpus quality and token efficiency are not backed by specific studies or data, and the video is primarily based on anecdotal experience. The sources cited are mostly links to the creator’s own content and promotional materials, with no external scientific references. The adéquation between title and content is good, as the video delivers exactly what it promises. The presence of promotional segments is noted but does not detract from the core value. Overall, the video is a useful tutorial for those seeking to enhance their AI interactions, but it lacks depth in terms of evidence-based justification.

159 words

Title / Content Match

The title accurately reflects the content: it promises a method to transform AI outputs from plain text to visually rich HTML documents, which is exactly what is demonstrated.

Quality & Reliability

7/10

The video provides a practical, reproducible technique (HTML formatting for AI responses) with clear demonstrations and references to a known AI researcher (Karpathy). However, it lacks formal citations or scientific evidence for the claimed benefits (e.g., token efficiency, better corpus quality), and the promotional elements are present.

Chapters

Cited Sources

Concurring Sources

Dissenting Sources

  • No direct discordant sources found — The video does not present conflicting information, but the claims about token efficiency and corpus quality are not substantiated by external sources.

External References

Contribution & Novelties

The video introduces a simple yet effective technique of requesting HTML-formatted responses from AI models, which significantly improves readability and user experience. It adapts a developer-focused tip for a broader audience, providing practical examples for various use cases. The approach is novel in its emphasis on presentation as a key factor in AI interaction quality.

Pour aller plus loin :

  • HTML — Official documentation on HTML, the markup language used in the technique.
  • Markdown — The original Markdown specification, often used for plain text formatting, which the video contrasts with HTML.
  • Prompt engineering — Wikipedia article on prompt engineering, the broader field of crafting effective instructions for AI models.
  • Token efficiency — Wikipedia article on tokens in machine learning, relevant to the claim about token savings.

126 words

Radar Profile

The radar profile shows balanced scores across information quantity, quality, technical level, and reliability, with a slight emphasis on practical applicability. The video is strong in providing actionable advice but weaker in scientific rigor and evidence.

Reliability 6/10