Le second cerveau IA de Karpathy est une impasse (+ ce qu'il devrait faire à la place)

Le second cerveau IA de Karpathy est une impasse (+ ce qu'il devrait faire à la place)

by Eliott Meunier

Humanities, Social Sciences & Thought AI — News & Applications UYComputer scienceUYQArtificial intelligence (AI)
🎙 Eliott Meunier 👥 51K 📅 April 15, 2026 ⏱ 24 min 👁 6K 📄 expert opinion 🧭 2026-08-27
Available in: English (current) Français

Keywords

ontologieIPCRAZettelkastenknowledge graphtaxonomie

Summary

The video critiques the viral ‘LLM Knowledge Bases’ trend, particularly the approach popularized by Andrej Karpathy, which involves using AI to automatically generate a personal wiki from raw sources. The author argues that this approach lacks a solid ontological foundation, leading to systems that are not sustainable over time. He explains the concept of ontology and reviews three classical types: hierarchical taxonomy, faceted classification, and knowledge graphs, highlighting their strengths and weaknesses for personal use. He emphasizes that the most dangerous ontology is the one not consciously chosen, and that Karpathy’s implicit ontology is a thematic wiki, which will become noisy and unmanageable as it grows. The author then presents two ontologies he recommends: IPCRA (Inbox, Projects, Hats, Resources, Archive), which organizes notes by utility and context rather than theme, and Zettelkasten, which focuses on atomic concepts and connections to foster creativity. He explains how these can be combined and how AI can enhance them by adding layers of connection and acting as a thinking partner. The video concludes with practical advice on implementing these systems, emphasizing the importance of choosing an ontology deliberately.

184 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides valuable insights into knowledge management, clearly explaining the concept of ontology and its importance. The argumentation is solid, as the author systematically critiques the viral approach by pointing out its lack of explicit ontology and potential long-term issues. He supports his claims with concrete examples and personal experience, making the reasoning persuasive. However, the argumentation relies heavily on anecdotal evidence and personal authority rather than empirical data or external studies, which slightly weakens the overall rigor.

Scientific Rigor, Source Quality, Title Accuracy

The video demonstrates a good level of scientific rigor in its conceptual explanations, referencing established frameworks like Zettelkasten and Arthur Koestler’s holarchies. However, it lacks direct citations to academic sources or studies, and the critique of Karpathy’s approach is based on personal opinion rather than systematic analysis. The title accurately reflects the content, and the video stays on topic throughout. The description provides links to Obsidian and a bootcamp, but these are not used as sources for the claims made.

174 words

Title / Content Match

The title accurately reflects the content: the video critiques Karpathy's LLM knowledge base approach and proposes alternative ontologies.

Quality & Reliability

7/10

The video presents a well-structured expert opinion on knowledge management ontologies, drawing on established frameworks (IPCRA, Zettelkasten) and referencing Arthur Koestler's holarchies. However, it lacks empirical data or citations to scientific studies, and the critique of Karpathy's approach is based on personal experience rather than systematic analysis.

Chapters

Cited Sources

  • Obsidian — Mentioned as the tool used for the knowledge base.
  • Prisme One Bootcamp — Promoted in the description as a course to build a Second Brain AI.

Concurring Sources

Dissenting Sources

  • Karpathy's approach — The video criticizes Karpathy's method of using AI to automatically organize notes, arguing it lacks a deliberate ontology and will not scale.

Contribution & Novelties

The video offers a fresh perspective on the recent trend of AI-generated personal knowledge bases, emphasizing the importance of ontology. It provides a clear comparison of existing ontologies and introduces IPCRA as a practical solution for personal knowledge management, which is a valuable contribution. The discussion on combining IPCRA with Zettelkasten for both project execution and creativity is particularly insightful.

Pour aller plus loin :

  • Zettelkasten method — A note-taking method that emphasizes atomic notes and connections, directly relevant to the video’s discussion.
  • Ontology (information science) — Provides a formal definition and context for the concept of ontology in information science.
  • Arthur Koestler’s holarchy — Explains the concept of holarchy, which the video references in the context of hierarchical taxonomies.

120 words

Radar Profile

The radar profile shows high scores in information quantity and quality, with moderate technical level and reliability. This indicates a well-structured and informative video, but with room for improvement in technical depth and source rigor.

Reliability 6/10