AI News: This Video Model Has Everyone Freaked Out!

AI News: This Video Model Has Everyone Freaked Out!

by Matt Wolfe

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 Matt Wolfe 👥 1.0M 📅 February 13, 2026 ⏱ 30 min 👁 113K 📄 news review 🧭 2026-08-28
Available in: English (current) Français

Keywords

Seedance 2.0AI videoGPT-5.3 Codex SparkGemini 3 Deep ThinkGLM-5

Summary

Matt Wolfe presents a weekly AI news roundup, highlighting the release of ByteDance’s Seedance 2.0 video generation model, which supports text, image, audio, and video inputs, producing highly realistic 15-second clips with synchronized audio. He showcases several impressive demos, including a UGC product video, a Waffle House clip, and a Lord of the Rings parody, noting the model’s ability to handle copyrighted characters and its unavailability in the US until February 24th. The video also covers Kling 3.0, now available on Leonardo AI, and Alibaba’s Qwen-Image-2.0, which the host tests with mixed results. Google’s Gemini 3 Deep Think model, available only to Ultra subscribers, shows state-of-the-art performance on benchmarks like ARC-AGI-2 and Humanity’s Last Exam, including gold-medal-level results on physics and chemistry olympiads. OpenAI introduced GPT-5.3 Codex Spark, using Cerebras chips for rapid inference, which the host demonstrates by generating a playable game in 50 seconds. Meta launched features to animate profile pictures and control Threads feeds via ‘Dear Algo’ posts. Z.ai released the open-source GLM-5 model, which approaches frontier performance on several benchmarks and was used to autonomously build a Game Boy Advance emulator. MiniMax M2.5 offers a highly cost-effective model for agentic tasks. The video concludes with rapid-fire news: OpenAI testing ads in ChatGPT, a viral Skarsgård ad that OpenAI denies, a delay of OpenAI’s hardware device to 2027, and deprecation of older models like GPT-4o.

228 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides substantial value by aggregating and demonstrating the latest AI developments, particularly in video generation and large language models. The host’s hands-on testing of GPT-5.3 Codex Spark and Qwen-Image-2.0 adds practical insight, and the inclusion of benchmark comparisons helps contextualize model performance. The argumentation is generally balanced, acknowledging both the impressive capabilities and potential concerns, such as copyright issues with Chinese models and the slippery slope of advertising in ChatGPT. However, the analysis is largely surface-level, relying on vendor claims and social media demos without deep technical scrutiny.

Scientific Rigor, Source Quality, Title Accuracy

The video demonstrates good sourcing by providing direct links to official announcements, blog posts, and social media demos for each major story. The host clearly distinguishes between official information and community-generated content. The title accurately reflects the content, focusing on the viral Seedance 2.0 model, and the video delivers on that promise. However, the reliance on unverified social media clips and the lack of independent verification of benchmark claims slightly reduce the overall rigor. The ad segment is clearly marked as sponsored and does not detract from the content.

194 words

Title / Content Match

The title accurately reflects the content, focusing on a viral video model and other AI news.

Quality & Reliability

7/10

The video is a weekly news roundup with direct links to official announcements and demos. The host provides personal testing and commentary, but the content is not peer-reviewed and relies on vendor claims and social media demos.

Chapters

Cited Sources

Concurring Sources

  • Seedance 2.0 Launch — Official announcement confirming the model's capabilities and release.
  • Gemini 3 Deep Think — Google's official blog post with benchmark results.
  • GPT-5.3 Codex Spark — OpenAI's official announcement of the model.

Dissenting Sources

  • OpenAI Device Delayed — The video mentions the delay, but the initial rumor of a Super Bowl ad was denied by OpenAI, creating a discrepancy.

External References

Contribution & Novelties

This video provides a timely and comprehensive overview of the latest AI developments, particularly highlighting the rapid progress in video generation with Seedance 2.0 and the increasing speed and capability of LLMs like GPT-5.3 Codex Spark and GLM-5. It offers practical demonstrations and context for understanding the implications of these advances, such as the potential for realistic deepfakes and the democratization of AI through open-source models.

Pour aller plus loin :

  • Video generation models — Overview of the field and its evolution.
  • Cerebras Systems — Information on the company behind the chips used in GPT-5.3 Codex Spark.
  • Humanity’s Last Exam — Details on the benchmark mentioned in the video.
  • Agentic AI — Concept of AI agents that autonomously plan and execute tasks, as demonstrated by GLM-5.

126 words

Radar Profile

The radar profile shows high scores in quantity of information and technical level, reflecting the video's comprehensive coverage and hands-on demonstrations. Quality and reliability are slightly lower due to reliance on vendor claims and social media demos, but still solid for a news roundup.

Reliability 7/10

💬 Très positif. Sur les 30 commentaires analysés, la majorité exprime de l'enthousiasme pour les démonstrations de Seedance 2.0 et la qualité des informations, avec quelques inquiétudes sur la difficulté croissante de distinguer le réel de l'IA.