Peut-on écrire un bon livre avec l'IA ? J'ai fait le test

Peut-on écrire un bon livre avec l'IA ? J'ai fait le test

Can you write a good book with AI? I tested it

by Erwan | IA · View the channel on PSI

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 Erwan | IA 👥 17K 📅 October 4, 2026 ⏱ 12 min 👁 20 📄 original study 🧭 2026-10-04
Available in: English (current) Français

Keywords

AI writingbook generationGPT-6Claudehuman creativity

Summary

The video presents a personal experiment where the author uses AI, specifically GPT-6 Astra in Codex, to autonomously produce a 143-page philosophy book titled ‘Esprits libres’. He details the entire workflow: creation of a custom ‘skill’ (a set of instructions), the AI performing research, planning, writing, revising, and formatting, followed by an independent audit and revision by Claude Opus 5.5. The final output includes a professional cover, multiple file formats, and a cohesive structure across 12 philosophical thinkers. The author reports a production time of 3 hours 20 minutes and an API cost of $101. He showcases excerpts and explains that while the book has the appearance of a real creation, the core creative decisions and intention remain irreplaceably human. The video concludes with reflections on the changing nature of creative work, the importance of human intent, and the societal debate around AI-generated art and literature.

146 words

Critical Evaluation

Value of the Information & Strength of the Argument

The primary value lies in providing a concrete, replicable methodology for AI-assisted book production, including cost breakdowns and technical insights (e.g., using a skill to manage long-context tasks). The argumentation is balanced: the author acknowledges the impressive quality of the output but firmly maintains that AI cannot rival human genius in terms of emotional depth and original intent. He supports this by emphasizing that the book, while well-constructed, lacks the ‘asperities’ (rough edges) of a truly human creation. The reasoning is logically sound and avoids hype, presenting a cautious yet nuanced view of AI’s capabilities in creative domains.

Scientific Rigor, Source Quality, Title Accuracy

The experiment is conducted with a clear and transparent methodology, including the creation of a detailed skill, autonomous AI operation, and a second-model audit. The author openly states the limitations and provides the generated book free of charge for scrutiny. Sources are limited to his own substack and the book’s download link, which are relevant and credible within the context of a practical demonstration. The title accurately matches the content. However, the reliance on personal experience without external benchmarks or peer feedback reduces the scientific rigor to an observational study rather than a controlled experiment. Comment trends are not analyzed due to absent data.

217 words

Title / Content Match

The title accurately reflects the content: a personal, hands-on test of whether AI can produce a good book.

Quality & Reliability

8/10

Highly transparent, detailed account of an experiment with clear methodology, costs, and limitations. However, it is a single-case study with no external validation, so the reliability is good but not exhaustive.

Chapters

Cited Sources

Contribution & Novelties

This video contributes a recent, practical case study on the current state of AI in long-form creative writing, specifically demonstrating an end-to-end autonomous pipeline with cost and time metrics. It offers a reproducible framework (the ‘skill’) that others could adapt, and it raises timely ethical and philosophical questions about authorship and creativity. For deeper exploration, see:

Pour aller plus loin :

  • Large language model — Essential background on the technology driving the experiment.
  • Artificial intelligence art — Context on AI in creative fields, including literature and design.
  • Human–AI collaboration — Relevant to the discussion of delegating tasks vs. preserving human oversight.
  • Copyright and artificial intelligence — Addresses legal and ethical implications of AI-generated works.

114 words

Radar Profile

The profile shows a balanced performance with high scores in information quality and reliability, moderate technical depth, and slightly lower quantity due to the short duration. The strong transparency and clear communication align with the high quality and reliability scores, while the technical level indicates an intermediate audience.

Reliability 8/10