J'ai demandé la même app à GPT-6 Astra et Fable 5.1 (c'est décevant...)

J'ai demandé la même app à GPT-6 Astra et Fable 5.1 (c'est décevant...)

I asked the same app from GPT-6 Astra and Fable 5.1 (it's disappointing...)

Title history
  • J'ai demandé la même app à GPT-6 Astra et Fable 5.1 (c'est décevant...) — observed on 20261110-0251 UTC (YouTube)

by Julien Massey · View the channel on PSI

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 Julien Massey 👥 39K 📅 September 18, 2026 ⏱ 20 min 👁 7K 📄 expert opinion 🧭 2026-11-10
Available in: English (current) Français

Keywords

AI comparisonGPT-6Fable 5.1app developmentautomation

Summary

The video presents a hands-on comparison between two AI models, GPT-6 Astra and Fable 5.1, across five tasks: a 3D solar system simulation, a Google Sheets clone, a voice agent for garages, invoice automation, and a quote generator. The creator reveals results gradually and scores each test. He concludes that Fable 5.1 is more consistent in design and one-shot generation, while GPT-6 excels in speed, automation, and integration with Google tools. He suggests GPT-6 for corporate automation and Fable for design-heavy one-shot projects. The video includes self-promotional segments for the creator’s agency, but these do not affect the core comparison. Each test is described with specific details about the output, including design quality, functionality, and ease of use. The final verdict contrasts regular performance (Fable) with high amplitude (GPT-6), implying GPT-6 can be brilliant or disappointing depending on the task.

140 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides practical insights from a user perspective, illustrating how each AI performs in real-world scenarios. The argumentation is based on subjective evaluation of the outputs, with the creator assigning points based on his own preferences. While this offers a relatable perspective, it lacks statistical rigor or controlled experimental design. The creator mentions independent benchmarks early on but does not cite specific sources, and his conclusions are drawn from a limited number of tests. The value lies in the concrete examples and the comparative framework, which can help viewers understand trade-offs between the models, but the evidence is anecdotal and potentially biased by the creator’s professional interests.

Scientific Rigor, Source Quality, Title Accuracy

The video does not cite any scientific sources, relying solely on the creator’s personal tests and impressions. The benchmarks mentioned are not linked or detailed, reducing the reproducibility and scientific credibility. The title accurately describes the core activity—asking the same app to both models—and the sentiment of disappointment is expressed for some results. However, the presence of self-promotional segments for the creator’s agency (around 6:00 and 19:40) introduces a commercial bias, though it does not directly invalidate the comparisons. The lack of objective metrics and source citations significantly weakens the scientific rigor. No comments were provided to analyze public reception.

223 words

Title / Content Match

The title accurately reflects the content: the creator gives the same prompt to both AI models and compares the results.

Quality & Reliability

5/10

Subjective comparison with clear promotional bias; lacks controlled experiments and source citations.

Chapters

Contribution & Novelties

The video contributes a practical, hands-on comparison of two leading AI models in realistic tasks, which is useful for practitioners choosing between them. It highlights differences in design quality, automation capabilities, and integration options, offering actionable insights for business use. However, the methodology is not novel and the conclusions are subjective.

Pour aller plus loin :

  • Large language model — Foundational concept for understanding these AI models.
  • Prompt engineering — Key technique for optimizing AI outputs, relevant to the tests.
  • Automation — Related to the invoice and payment automation tasks showcased.

91 words

Radar Profile

The scores show a moderate quantity of information but lower quality and reliability due to subjective testing. The technical level is moderate, making the content accessible to a general audience, but the overall fiabilité is compromised by lack of sources and promotional bias.

Reliability 4/10