Google aplastó a todos con Gemini 3, la mejor IA de 2025

The video provides a valuable hands-on demonstration of Gemini 3’s capabilities, particularly in generating code and animations from simple prompts. The presenter’s approach of testing with real-world tasks, such as creating an SVG animation of a VPN and an interactive geometry demonstration, offers concrete evidence of the model’s performance. The transparency about failures (e.g., the map not being accurate) adds credibility. However, the evaluation is limited by the short time since release; the presenter admits he hasn’t tested deeply. The video relies heavily on Google’s promotional claims, such as benchmark scores and user numbers, without independent verification. The comparison with Claude is useful but not exhaustive. The argumentation is generally sound, but the tone is enthusiastic and somewhat promotional, which may bias the assessment. The sources cited are limited to EDteam’s own links, with no external references to Google’s official documentation or research papers. The title is slightly sensationalist but the content largely supports it. Overall, the video is informative for a general audience interested in AI developments, but it lacks critical depth and rigorous sourcing.

The title is somewhat sensationalist ('aplastó a todos') but the video does demonstrate Gemini 3's capabilities and compares it with Claude, so it is largely accurate.

6/10

The video is a hands-on review of Gemini 3, showing real-time tests and comparisons with Claude. The presenter is transparent about limitations (e.g., not testing deeply, some failures). However, the content is promotional in tone, relies on Google's claims, and lacks independent verification or citations. The benchmarks mentioned are from Google's announcement, not independently verified.

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

• 00:00Introduction: Black Friday promo, announcement of Gemini 3 release on Nov 18, 2025.
• 02:00Overview of Gemini 3's promised features: advanced reasoning, multimodal, non-condescending.
• 05:00Discussion of Google's user numbers and developer statistics.
• 07:00Benchmark scores for Gemini 3, including ARC-AGI and deepsync.
• 10:00First test: creating an SVG animation explaining VPNs.
• 13:00Iterating on the VPN animation with more realistic nodes and a map.
• 16:00Test with hand-drawn diagram to create 3D animation using Three.js.
• 19:00Comparison with Claude on the same VPN animation task.
• 22:00Test: interactive animation for triangle angle sum, first attempt fails.
• 25:00Improved prompt leads to successful demonstration using 'tearing corners' method.

• EDteam free courses — Mentioned as a resource for learning technology.
• EDteam courses — General link to EDteam's course catalog.
• EDteam scholarships — Mentioned for students seeking scholarships.
• EDteam LinkedIn — Social media link.
• EDteam Instagram — Social media link.
• EDteam premium — Mentioned for premium access.
• EDteam instructors — Mentioned for those interested in teaching.
• EDteam TikTok — Social media link.

• Google AI blog — Official source for Gemini 3 announcements and claims.
• Gemini 3 official page — Official page for Gemini models.

• OpenAI