Claude Fable 5 : qui gagne, qui perd, et quoi faire avant le 23 juin

The video provides a compelling and detailed analysis of Claude Fable 5, based on the author’s personal testing and financial reasoning. The author demonstrates a strong understanding of the technical and economic aspects of AI models, and his argument about the shift to usage-based pricing is well-supported by the evidence he presents. He cites official Anthropic sources and a tweet by Andrej Karpathy, which adds credibility. However, the analysis is largely subjective, relying on personal experience and speculation about Anthropic’s motives. The author’s prediction about the permanence of usage-based pricing is plausible but not certain. The video also includes a promotional segment for his Patreon, which is clearly disclosed. Overall, the video is informative and thought-provoking, but viewers should be aware of the author’s personal perspective and potential biases. The title accurately reflects the content, and the video is well-structured with clear sections. The main strengths are the detailed cost analysis and the strategic insights. The main weaknesses are the lack of independent verification of the claims and the speculative nature of some conclusions.

The title accurately reflects the content, focusing on the winners and losers of Claude Fable 5 and what to do before June 23.

7/10

The video provides a detailed personal analysis of Claude Fable 5, including cost measurements and strategic implications. The author is transparent about his methods and sources, but the analysis is largely based on personal experience and speculation about Anthropic's business strategy. The sources cited are official Anthropic pages and a tweet by Karpathy, which are reliable, but the interpretation is subjective.

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

• 00:00Introduction: The author recalls his April prediction about price multiplier and announces the launch of Claude Fable 5.
• 02:00Pricing details: Fable 5 costs $10/$50 per million tokens, double Opus 4.8, and is included in subscriptions until June 22.
• 05:00Personal testing: The author rebranches his missions on Fable 5 and notes a significant performance improvement.
• 10:00Cost measurement: A 2.5-hour mission consumed 800k tokens and 45M context tokens, costing nearly $200.
• 15:00Limitations: The author lists three limitations: price, consumption, and guardrails.
• 20:00Cost management tips: The author advises requesting a plan before execution and treating context as precious fuel.
• 25:00Industry shift: The author discusses the move from buying answers to buying missions, citing Stripe case and Anthropic's blog.
• 30:00June 23 change: Fable 5 will be removed from subscriptions and require usage credits, signaling a permanent shift to usage-based pricing.
• 33:20Conclusion: The author predicts Fable 5 will remain usage-based and advises viewers to prepare.

• Anthropic Claude Fable page — Official pricing and details for Fable 5, including the Stripe case and subscription inclusion until June 22.
• Anthropic Institute: When AI builds itself — Blog post from June 4 discussing autonomous task duration doubling every ~4 months and 80%+ of internal code written by Claude.
• Andrej Karpathy's tweet — Karpathy's reaction to Fable 5 launch, noting a 'step change' and more ambitious tasks.

• Anthropic Claude Fable page — Confirms pricing and subscription inclusion details.
• Anthropic Institute: When AI builds itself — Supports the claim about increasing task durations and AI-written code.