Pourquoi ton AIOS est malade (et comment le soigner)

Pourquoi ton AIOS est malade (et comment le soigner)

Why your AIOS is sick (and how to fix it)

Title history
  • Pourquoi ton AIOS est malade (et comment le soigner) — observed on 20261111-2334 UTC (YouTube)

by ThomRichard · View the channel on PSI

Applied Sciences & Engineering Computing & Cybersecurity UComputing and Information TechnologyULOperating systems
🎙 ThomRichard 👥 6K 📅 September 22, 2026 ⏱ 14 min 👁 579 📄 tutorial 🧭 2026-11-11
Available in: English (current) Français

Keywords

AIOSsingle source of truthAI maintenanceClaude skillsproductivity

Summary

The video addresses common malfunctions in personal AI operating systems (AIOS), the user’s AI-driven productivity setup. The creator, Thomas, describes two main symptom families: the AIOS no longer integrates or retrieves correct information, and the AIOS no longer performs tasks according to expectations. For each, he proposes structural fixes (capture points, consistent ontology like IPCRA, single source of truth, functional map, decision trees) and maintenance rituals (session closure, audits, level-up to formalize repeated actions, decision journal). He emphasizes treating recurring problems rather than rebuilding from scratch, and cautions against over-engineering: one should start with a minimum viable product and add features only when a real need and value justify it. The advice relies on personal experience coaching around 200 users, and no formal data or controlled evidence is presented. Nevertheless, the framework is practical, coherent, and useful for users of AI assistants like Claude, ChatGPT, or Cursor-based workflows.

148 words

Critical Evaluation

Value of the Information & Strength of the Argument

The main value lies in offering a structured diagnosis framework for personal AI systems. The author distinguishes structural problems (how information is captured, classified, and sourced) from maintenance problems (rituals like audits and decision logs), which is a genuinely useful mental model. The argumentation is coherent and illustrated with concrete examples such as meeting minutes, outdated prices, and inconsistent tone in emails. However, the claims are supported only by personal experience and testimonials from a private coaching community; there is no empirical evidence, benchmarking, or independent validation. The ’level-up’ ritual borrows from a named source (Nate Earth) but without a proper reference. Overall, the video is persuasive for practitioners seeking quick fixes, but it should not be treated as scientific literature.

Scientific Rigor, Source Quality, Title Accuracy

Scientific rigor is limited. The video involves practical heuristics rather than formal research, and the creator does not cite academic studies. Mentions of IPCRA, Nate Earth, and the concept of single source of truth are not backed by external links or verifiable citations. The title is well aligned with the content, offering a clear and accurate promise. The description includes links to the creator’s own masterclass, newsletter, and community, as well as two short explainer videos about AIOS and Claude skills, but these are primarily educational resources on the same channel rather than independent sources. No contrary evidence or alternative viewpoints are discussed.

239 words

Title / Content Match

The title accurately reflects the content: the video diagnoses sources of AIOS malfunction and proposes remedies, without overpromising or straying from the topic.

Quality & Reliability

6/10

The content is practical and internally coherent, based on the creator's personal experience with a small audience, but it lacks formal evidence, peer-reviewed sources, or quantitative validation. The advice is plausible and actionable, yet remains anecdotal and somewhat self-promotional.

Chapters

Cited Sources

External References

Contribution & Novelties

The novelty lies in applying a medical metaphor to AI system maintenance, offering a reusable diagnostic checklist that separates structural issues from recurring ritual failures. It encourages users to avoid rebuilding an AIOS from scratch, which is a common and tempting but unproductive habit. The video provides several concrete techniques (capture points, single source of truth, decision journal, level-up skill, 10-second ontology test) that are not commonly described in such a compact way.

Pour aller plus loin :

  • Single source of truth — Key principle for reliable data management, directly relevant to the video’s advice on avoiding duplicate or obsolete information.
  • Minimum viable product — The creator explicitly recommends starting with an MVP for an AIOS; this Wikipedia article explains the concept and its trade-offs.
  • Model Context Protocol (MCP) — Mentioned indirectly through ‘connexions / MCP’; this protocol standardizes how AI systems connect to tools and data, supporting the video’s emphasis on cleaner integrations.

154 words

Radar Profile

The radar profile shows moderate scores across all dimensions, with quantity of information slightly higher than quality and technical level. This reflects a practical, experience-based video that offers useful material but with limited scientific depth and rigor. The balance suggests the content is most valuable as an introductory practical guide rather than a technical reference.

Reliability 5/10