J’avais (encore) tort… Google, c’est vraiment fort en fait

J’avais (encore) tort… Google, c’est vraiment fort en fait

I was wrong (again)... Google is actually really strong

Title history
  • J’avais (encore) tort… Google, c’est vraiment fort en fait — observed on 20261101-0039 UTC (YouTube)
🎙 Arthur (DevArt) 👥 19K 📅 September 8, 2026 ⏱ 17 min 👁 19K 📄 expert opinion 🧭 2026-11-01
Available in: English (current) Français

Keywords

Gemini 3.8 FlashGoogle AILLM comparisoncodingcost-effective AI

Summary

The video is a practical evaluation of Google’s Gemini 3.8 Flash model for software development. The author tests the model on four different projects: an iPhone emulator, a 3D space interactive experience, an airspace live scanner, and a voice transcription app. He highlights the model’s speed, low cost, and impressive coding capabilities, comparing it favorably to frontier models from Anthropic and OpenAI. He also demonstrates the use of Antigravity CLI for an efficient workflow. The video concludes that Gemini Flash offers exceptional value for most coding tasks, challenging the notion that expensive frontier models are necessary. The creator provides pricing details and a transparent cost-benefit analysis, making a compelling case for budget-conscious developers. The demonstrations are visually impressive and show the model’s capability to generate complex interactive applications from simple prompts. The overall tone is enthusiastic but backed by concrete examples and a willingness to share code repositories.

148 words

Critical Evaluation

Value of the Information & Strength of the Argument

The author argues that Gemini Flash is a cost-effective alternative to expensive frontier models. He supports this with concrete examples and detailed pricing comparisons. He openly discusses limitations, such as hitting API rate limits, but overall the argument is persuasive. The main strength is the practical evidence; however, the lack of objective benchmarks limits the robustness. The video effectively builds a case by showing that even a lower-cost model can produce high-quality results for a majority of use cases, thus challenging the necessity of premium models.

Scientific Rigor, Source Quality, Title Accuracy

The video cites official Google sources and provides transparent pricing from OpenRouter. The title matches the content as it reflects the author’s change of opinion. However, the evaluation is anecdotal and not a scientific study. The author does not delve into potential biases and his personal experience might not generalize. The sources include direct links to the official model page and the CLI tool, which adds credibility. The title is slightly sensational but accurately represents the video’s message.

179 words

Title / Content Match

The title reflects the content well: the author admits being surprised by the capabilities of Gemini Flash, which is exactly what the video demonstrates.

Quality & Reliability

8/10

The creator provides practical tests and transparent cost details, referencing official Google documentation. However, the evaluation is based on personal experience rather than a systematic benchmark, so some subjectivity is present.

Chapters

Cited Sources

Concurring Sources

External References

Contribution & Novelties

This video provides a hands-on demonstration of Gemini Flash’s capabilities beyond simple benchmarks, showing real-world applications. It also offers a cost-benefit analysis for developers. The originality lies in the comparative pricing data and the practical test scenarios that are not commonly found in typical model reviews.

Pour aller plus loin :

96 words

Radar Profile

The score profile shows a balanced performance with slightly higher reliability and information quantity, while the technical depth is moderate. This reflects a practical, hands-on review that emphasizes real-world applicability over academic rigor.

Reliability 8/10